While major stock indexes were hitting all-time highs and CCaaS vendors were issuing press releases about their “transformative AI capabilities,” I was calling their bluff and predicting a category-wide repricing.

The market has now validated that thesis. What follows is the framework I used and what it says about where the industry actually goes from here.

The Call

In Q1 2023, ChatGPT hype was peaking.

NICE announced ChatGPT integration. Anyone with basic technical literacy could see it was pure marketing.

Five9 teased “game-changing” GPT-powered solutions. RingCentral talked AI on every earnings call. Enterprise Connect was saturated with AI announcements.

Analysts upgraded these stocks on the assumption that generative AI would accelerate cloud migration and increase platform stickiness.

I said to avoid them all. Except Zoom.

The industry laughed and shrugged them off, meanwhile I moved my entire business to ZoomCX in Q3 of 2023 and publicly called my shot.

Here’s how that played out:

NICE’s market cap fell from roughly $16 billion to $7 billion. Zoom’s increased from roughly $19 billion to $27 billion over the same period.

Digest that for a moment.

Subscribe now

The Framework: Four Structural Constraints

I didn’t call this on vibes. I used EthOS.

EthOS is the internal operating system I’ve developed, similar to Koch’s Market-Based Management, but we also apply it externally to evaluate structural risk, capital allocation, and category durability.

When I evaluated CCaaS vendors through that lens, four failures were obvious.

1. Technical Debt from Roll-Up Strategies

When customers publicly refer to NICE CXone as “CX12,” they are accurately describing a problem.

NICE, Five9, RingCentral, 8x8 have all grown through acquisition. That means years of bolting purchased companies together rather than building coherent architecture.

With that level of integration debt, you cannot “add AI.” Every feature becomes a negotiation with past compromises.

Press releases said “AI-powered”.

Engineering reality said unresolved internal dependencies.

Zoom followed a different path. They built infrastructure early, hardened it during the pandemic, and then focused on core CX primitives built on a single architecture. As a result, their development velocity was unmatched.

2. Market Constraints on AI Adoption

While vendors promise AI transformation on investor calls, enterprise reality is governed by different physics.

Healthcare deployments require compliance review and regulatory approval. Financial services procurement cycles routinely exceed a year. Unionized contact centers require negotiation before automation. Fortune 500 change management operates on quarterly cycles.

Actual AI deployment velocity in contact centers is far slower than the timelines implied in earnings calls.

Vendors were selling outcomes their customers could not operationally reach and recognizing revenue on sales cycles that were lengthening.

3. Commoditized AI Eliminated the Moat

This is where the market fundamentally misread the situation.

When a CCaaS vendor announces LLM integration, they are announcing access to the same API available to any startup.

OpenAI, Anthropic, and Google Cloud are commodity suppliers. The API is the API. There is no proprietary intelligence advantage.

For years, incumbents sold complexity as differentiation. Routing logic, IVR tuning, workforce optimization. Generative AI collapses that complexity.

When the hard part becomes easy, pricing power disappears.

4. M&A Exhaustion

The growth model stopped working.

While NICE was busy attributing their YoY growth to AI, I seemed to be the only one doing the math that a lion’s share of it was coming from acquisitions that they were burying, and that they were running out of options for future deals.

The constraints were straightforward:

• The market is too concentrated for meaningful acquisitions

• Equity currency has been destroyed

• Integration capacity is exhausted

They cannot acquire their way forward. Organic growth requires architectural coherence they do not have.

Why Zoom Thrived

Zoom’s outcome was not accidental.

They built CCaaS as a modern system, not as an accumulation of legacy products. No acquisition cleanup. One architecture.

When Zoom integrates AI, it is integrated into a platform rather than layered onto fragmented systems.

Eric Yuan built a platform. The others assembled portfolios.

That distinction is invisible on a balance sheet, but obvious in a constraint analysis. And it’s the difference between -77% and +40%.

The Final Dimension: Leadership

There is an additional layer the numbers don’t capture.

Executives declaring leadership while announcing integrations that did not exist. CEOs repackaging open-source components as proprietary breakthroughs. CTOs celebrating “multi-model strategies” as if that somehow constitutes strategy.

These signals matter. They reflect internal confusion and assume external ignorance.

By contrast, Eric Yuan embodies humility while emphasizing reliability, customer experience, and execution discipline. That posture propagates through his entire organization and creates resilience.

Leadership does not override structural constraints. It determines how organizations behave once those constraints tighten.

What’s Next

The CCaaS valuation collapse was not an anomaly. It was a structural outcome that could be seen years in advance with the right lens.

The next one is already visible.

That is for another post.

• Voice IVA was overfunded, overhyped and did not automate away the industry like the YC kids wanted us to believe.

• “Agentic Orchestration” is still an abstract concept and demo-ware at best.

• Agent Assist is useful, but only for a small percentage of interactions that require depth or non-routine complexity.

• Automated QA has become a checkbox illusion of AI adoption, where the noise-to-value ratio is inverted.

• Post-call summaries matter little without intentional construction and automated data pushes into the CRM.

• Organizations still lack the know-how to implement and adopt AI with any measure of success, while most vendors continue pandering to the press.
  

So, in 2026, I'm putting the "Last Mile, First" by attacking the constraints to customer adoption.

Our thesis is Unified CX.

Customer experience breaks when strategy, tooling, deployment, and operations are split across vendors with misaligned incentives.

• Consultants advise without ownership.

• Integrators deploy and then leave.

• BPOs run agents but seldom own the full stack.

• Software vendors collect license fees whether the thing works or not.
  

Nobody owns outcomes. So nobody delivers them.

That's why I'm converging our separate AI-CX companies into one: InflectionCX

One system. One partner. End-to-end accountability.

We discover opportunity, design the solution, deploy the technology, and operate it in production with live Agents and AI ones -> converging people and AI into a single operating paradigm.

And, we’re running the opposite playbook of everyone else:

• We’re investing heavily in non-CX-facing systems that expand internal capacity and enable precision and scale: recruiting, learning management, and go-to-market.

• We’re putting Foundation over foundational models, ensuring that your data, context, governance, and processes exist before AI is deployed. If those aren't ready, neither is your AI.

• Decades of digital red-lining have left large parts of the country without residential high-speed internet. No internet means no remote work. So we’re putting call centers where they matter most: Rural America.
  

Which also means that next year, I get to go play in the dirt.

Subscribe now

This is the moment your automation dies. Your carefully mapped decision tree hits something it can’t handle. Day 31 on a 30-day return policy. A VIP customer using a competitor’s coupon code. The shipping delay that was technically your fault but not really.

Subscribe now

Call center agents know this moment well. The screen says one thing, but the situation demands judgment. So they escalate. And escalate. And suddenly your streamlined operation has layers of human override.

Here’s what’s actually happening: workflows encode rules, but businesses run on policy. Rules say “30 days means 30 days.” Policy says “keep valuable customers happy within reasonable bounds.”

That supervisor isn’t breaking the system when they approve the day-31 refund for your best customer. They’re applying the real policy that was too complex for your flowchart.

AI agents change this game entirely. Feed them policy, not workflows. Let them weigh customer lifetime value against return risk. Let them understand that “30 days” is a guideline, not a law of physics. Let them make supervisor-level calls at entry-level scale.

The companies winning tomorrow will have one agent handling what used to require three tiers of escalation. Not because the agent follows more complex workflows, but because it understands what the business actually wants to achieve.

Your workflows are just badly written policy. Time to give your agents the real thing.​​​​​​​​​​​​​​​​

Subscribe now

Then they use this Formula One intelligence machine to deliver compliance reports.

Right now, in some data center, a customer is recorded saying: “I’m switching to Competitor X because their platform actually remembers my preferences between sessions.” That exact quote, that specific product gap, that clearly stated switching trigger—it’s sitting in perfect digital clarity on a server nobody will ever access.

Down the hall, the product team is commissioning a $2 million study to understand why renewal rates are dropping.

In another recording, timestamped last Tuesday at 2:47 PM: “I would literally pay double if you just let me export my data to Excel without going through support.” The customer gave an exact price point. Specified the feature. Even used the word “literally.”

Meanwhile, the pricing team is hiring McKinsey to model willingness-to-pay.

The evidence accumulates by the second. A customer in Cleveland explains: “Your competitor’s checkout process takes two clicks. Yours takes seven. That’s why I’m leaving.” Another in Dallas details: “I spent three hours trying to integrate your API. Your documentation is wrong on line 347. Here’s the correct syntax…”

Precise. Actionable. Ignored.

These aren’t suggestions buried in survey responses. These are customers explaining, in their own words, at the exact moment of friction, what would make them stay, pay more, or recommend you to others. They troubleshoot your products. They reveal competitor advantages. They provide implementation fixes.

All recorded. All stored. All deleted from organizational consciousness the moment the call ends.

The perversity compounds. A telecommunications company spent $50 million upgrading their contact center technology last year. State-of-the-art speech analytics. Cloud storage. AI-powered transcription accurate to 97%.

They use it to verify agents said “thank you for calling” and properly disclosed the terms and conditions.

Meanwhile, 40,000 customers last month specifically mentioned their main competitor’s family plan pricing. 12,000 described a billing error that triggers every February. 8,000 explained the exact workflow that makes them consider switching.

The intelligence sits there, pristine and pointless, like a library where nobody reads.

Here’s what makes this criminal: the technology barrier disappeared five years ago. Storage costs nothing. Transcription is automatic. Search is instant. Analysis tools can process millions of conversations in minutes.

Companies solved every technical problem except the only one that matters: actually listening to what they’ve recorded.

Instead, they build parallel intelligence-gathering systems. They run focus groups to hear what customers already told them. They deploy surveys to learn what’s in their recordings. They hire consultants to discover what their contact center agents hear every day.

A retail bank I studied spent $8 million on customer research last year. In their contact recordings from the same period: 23,000 customers explaining exactly why they closed their accounts. 31,000 describing the specific mobile app features they need. 14,000 comparing their mortgage rates to specific competitors.

The research study took six months and surveyed 1,200 people. The recordings contained insights from 68,000 actual customers experiencing actual problems in real time.

The study sits in a PowerPoint deck. The recordings sit in cold storage.

This is the industrial deletion complex: a system perfectly designed to capture intelligence and systematically engineered to ignore it. Every company convinced they’re data-driven while drowning in the data that matters most.

Your customers aren’t hiding their needs. They’re screaming them into your recording systems.

You’re just paying $100 billion a year to hit mute.​​​​​​​​​​​​​​​​

Subscribe now

In the silence, I came to realize that the giant sucking sound we hear is the vacuum of leadership in the contact center space since AI showed up.

CX vendors today are mostly chasing trends and faux accolades. While analysts are stuck parsing their technical jargon and press releases, customers still can’t tell the difference between now and 2022 when ChatGPT arrived.

The market is a flash flood of AI tools that never considered the operator.

Transformation was promised yet more complexity was delivered.

The failure is of architecture, ownership, and intention.

What comes next isn’t a tech problem. We’re done debating whether AI can do more. It can and it will.

It’s now a reckoning of values where the real questions are:

- Who’s shaping what we actually want to get out of AI?

- Who’s willing to lead without a roadmap?

Industry transitions are never led by headline-chasers. They’re led by the people willing to rebuild the core of not just the tech stack, but the culture behind it.

Now we separate the operators from the opportunists so the future can be shaped by what we decide to stand for, not what vendors decide to ship.

It’s our work to do now.

Subscribe now

Designing our "Agentic" Coaching tool was a near existential crisis for me.

As in: laying on the floor of my office for hours pretending to "meditate." Six-mile runs that solved nothing. Bottomless cups of coffee that became a personality disorder. All the while secretly wishing someone would either make a decision or put me into retirement.

And then it hit.

The question wasn't "what do we build?" It was "what do we actually believe coaching should become?"

Oh shit. Anything but that.

More coffee. More meditation.

It's easy to build software that does what we've always done, just faster. The problem with AI is it makes it even easier to generate more noise under the guise of intelligence.

Call-by-call recommendations. Agent rankings. Performance analytics that look impressive in meetings. None of it actually changes anything. Just makes the chaos prettier.

Pig, meet lipstick.

So we went back to the beginning.

We were handing decision-making over to something that doesn't understand context. We were automating coaching without defining what coaching actually was.

AI doesn't just follow instructions. It amplifies assumptions. Every bias you don't address becomes a feature. Every value you don't define becomes a blind spot that scales.

We had to define Operational Excellence on a blank canvas. Not what the technology looks like but what the outcomes look like. What it feels like to be coached toward your potential instead of managed toward compliance.

We don't need guardrails. We need doctrine. We don't need more data. We need direction. We need to bridge the chasm between possibility and practice.

Define objectives. Map trajectories. Know your destination before you plot the pathway.

AI isn't the coach. It's the pathfinder.

But first, you have to know where you're going.

The tool that nearly broke me taught me this: Most companies are building solutions to problems they've never actually defined. They're automating workflows they've never questioned. Adding intelligence to systems that were fundamentally stupid to begin with.

Define your Ethos first, and let the tools follow.

At no other time have I seen such a massive disconnect between (most) vendors and customers. Vendors are focused on the press release for “What’s Next” yet customers want to know what they’re actually getting from what they’ve already bought. And absolutely no end-user that I talked to gave a shit about Agentic AI or Orchestration, yet the vendors roll on.

Let’s review:

• Talkdesk “shatters the industry” with more Orchestration buzzwords and declares a new software category. Adoption pending.

• CEO of 8x8 corrects Talkdesk by saying they released theirs first. Meanwhile, customers are wondering why the hell they need Zappier in a Tuxedo.

• NiCE - We changed the “i” to lowercase and added a half-cocked smiley emoji to our logo. And we’re just getting started.-Oh, and we have an Orchestration thingy too.

• Five9 launches another Agentic whatcha-ma-call-it and apparently only one person in the entire company can demo it because it’s not yet in production. Discern what you will.

• BPOs struggling to rationalize an AI strategy that differentiates them from everyone else that’s just using whatever their CCaaS is giving them or an OpenAI wrapper. One BPO in particular expressed their displeasure with one of my recent posts, and let me know that “facts don’t matter”. Facts do, in fact, matter.

• Another cohort of AI startups that have never stepped foot in a call center telling you how they’ve reinvented it.

Now, for some upside.

• If you weren’t aware, Zoom has a contact center solution and once the industry figures that out, the legacies are in trouble. Zoom was pretty chill about the new Voice IVA they launched this week, but if you want to see it in action, follow Tommy Hart. It’s in production and they’re happy to demonstrate the results for you.

• After spending time with Cisco, I’m even more bullish on their comeback. There’s an incredible amount of humility around the challenges they face in adapting their GTM strategy to align with the current market, as well as playing catch-up to the cloud. And their approach to AI doesn't suck.

Both Zoom and Cisco continue to lead the market in openness and customer engagement. They’ll both acknowledge what they do well, and what they don’t and are happy to partner in the best interests of their customers.

I value that.

My Agent Assist face-plant post buried the lede. We're a BPO. We torched this thing with three different clients before I asked the obvious:What are we actually trying to accomplish here?

Subscribe now

The Vanity Metrics Dashboard:

• Adoption rate: 87%! (agents clicking buttons)

• Queries per hour: 45! (agents clicking more buttons)

• Knowledge surfaced: 10K articles! (agents closing popups)

Cool metrics. Did it actually help anyone?

So We Asked Agents: "When Do You Actually Need Help?"

Their answers:

• Edge cases – 8% of calls

• New scenarios – 3%

• Policy conflicts – 4%

• "I know this but can't remember" – 6%

21% of calls. That’s the whole game.

We were force-feeding AI on 79% of calls they could handle blindfolded. They don’t need help updating addresses. They don’t need your refund policy script. They don’t need to be told how to greet a customer for the 10,000th time.

The Measurement Circus

What Actually Matters (on that 21%):

• Did agents give the right info? (not “did they click”)

• Did customers accept it? (not “was knowledge surfaced”)

• Did it prevent escalation? (not “adoption rate”)

• Did the customer call back? (not “query volume”)

The Real Problem:

Your Agent Assist vendor can’t measure what matters. They track their buttons. They don’t see that CSAT tanked, escalations spiked, or callbacks surged.

You bought a speedometer for a parked car.

Your Choice:

Impact 100% of calls by 1%? Or 21% of calls by 50%?

Agents already know. They’re clicking to dismiss popups while solving real issues with real skills.

Watch them scramble.

"Well, we focus on successful interactions..." "Our latest model has improved..." "Let me show you this perfect example instead..."

Subscribe now

Here's Why They Panic: Every AI fails the same way. Confident hallucination followed by agent cleanup. But vendors only screenshot the wins.

The Demo That Matters:

• Customer asks about policy exception

• AI confidently invents new policy

• Customer believes AI

• Agent spends 20 minutes undoing damage

• QA dings agent for long handle time

Your vendor has 1,000 of these. They're hoping you never ask.

The Agent Assist Version:

• Complex billing question surfaces

• AI suggests outdated procedure from 2022

• Agent knows it's wrong, follows it anyway (CYA)

• Customer gets incorrect resolution

• Callback next week doubles handle time

Same movie. Different theater.

The Vendor Blame Game: When I posted yesterday about AI failures, almost every vendor spit-up some version of “Your AI failure is most likely a data issue. Not an AI issue.”

Beautiful. They sold you a solution knowing your data was "a mess" but forgot to mention their AI can't handle reality.

Put this in your AI RFP:

• “What's your confidence scoring on wrong answers?"

• “How do agents report AI errors?"

• “If our data is messy, why did you promise results?"

• “Where's your kill switch?"

The Real Tell: If they claim their AI doesn't hallucinate, they're either lying or delusional. GPT-4 hallucinates. Claude hallucinates. Your $3M vendor definitely hallucinates.

The honest vendors say: "AI fails predictably. Here's how we handle it." The cowboys show more demos of perfect conversations.

What Actually Works:

• AI that knows when it doesn't know.

• Clear escalation paths.

• Agent override authority.

• Real-time accuracy monitoring.

• Measurement of actual help, not just usage.

But that doesn't demo well. So they show you store hours lookups while your reality involves Medicare Part D exceptions and promotional pricing conflicts.

Your Move: Ask for the disaster reel. Ask which calls it can't handle. Ask what agents do when it's wrong.

If they can't show you where AI fails, they can't show you where it works.

Better yet, ask your agents. They've named your AI "Shroomy" in group chat.

Maybe they should run your next vendor demo.

Six weeks later, agents were using it for exactly one thing: proving to their supervisor they'd "checked everything."

The Implementation Reality:

• Week 1: "This will revolutionize agent efficiency"

• Week 2: "Why isn't it surfacing the right answers?"

• Week 3: "Let's add more 20 documents. Make sure they’re structured just right.”

• Week 4: "Maybe we need better intent tagging"

• Week 6: “Um, can we get a refund on this thing?”

What Actually Happened:

The AI worked maybe worked ok. It revealed that our documentation was 87 different versions of "it depends."

Policies had exceptions. Procedures had workarounds. And the real information was still buried in someone’s Inbox and our Agents’ heads.

We didn't have knowledge to manage. We had years of band-aids documented as gospel.

The Beautiful Irony:

Agents were un-phased. Not because the tech failed, but because it proved what they'd always known: their job isn't following documentation. It's navigating the chaos between what we say and what we do.

The Uncomfortable Truth:

AI can't assist agents when your business rules are more like business suggestions. It surfaces exactly what you documented: institutionalized confusion.

Agents don’t need automated checklists and playbooks or their attendance policy to surface when a caller is running late for an appointment, because someone uploaded the company handbook into the same knowledge base as the “how to reschedule” FAQ.

The need precision support when it matters.

What if agent assist revealed the real problem isn't teaching AI your policies, but having policies worth teaching?

Your agents aren't worried about AI because they know something vendors don't: Most of their job is compensating for your company's inability to make clear decisions.

AI isn’t coming for their jobs. It can't even find the right policy.

They're safer than they've ever been.​​​​​​​​​​​​​​​​

It's a big day for me personally.

It's a big day for EndeavorCX, our company, and I believe for the contact center industry at large.

I've been fortunate enough in my 30-ish year career now to have seen and played a part of multiple market transitions.

Twenty years ago, we were moving from TDM to voice over IP.

Ten years ago, that migration from on-premise to the cloud and CCaaS and SaaS taking over the world began.

And it's no secret or surprise that the arrival of AI represents the largest opportunity for our industry to become what it can be.

The problem is that the vendors that led the last market transition to the cloud are failing to lead this one.

It's a different set of skills.

It's a different competence.

And instead, they've chosen hype over reality.

They've chosen to issue press releases instead of release notes and ship product.

They've chosen to change their messaging every quarter based on whatever the trend is for the day.

This quarter, it’s agentic AI.

They’ve chosen high per-agent-per-month pricing for bundles that are obtuse and don’t drive actual results.

And they’ve also chosen to lock up your data in proprietary formats and models, making it impossible for you to stay in control of your own destiny.

So today we change all that.

Today I’m calling this the great contact center rewrite.

It is unprecedented in the size, scope, scale, and depth of what we are doing.

Today we’re not releasing a new product.

We’re releasing a new architecture for the entire industry that puts you in control.

It puts the data where it belongs: in the center.

CCaaS is no longer the center of the universe. Your data is.

Over the next 12 weeks, we’re releasing 12 products, not a singular product.

We’re shipping 12 new solutions that will transform the way you think about AI, customer experience, agent performance, and operational excellence.

We’ll be announcing some amazing things that you’ve never seen before—things I’ve never seen before.

In my 25 years of building data teams and data products, we’re talking about knowledge bases that construct and write themselves, curate themselves, and deploy knowledge across the entire enterprise.

We’re talking about the ability to have a data scientist equivalent running on open source language models that can pinpoint issues inside of healthcare plans, as an example.

The constraint to realizing our potential in the industry is access to data—and two forms of data in particular.

One is the contact call detail records.

The other is the transcript.

Whether vendors want to tell you this or not, 80% of all use cases in the contact center AI space start with a transcript.

Call summary, sentiment analysis, identifying customer journeys, friction points, automating quality assurance—it all starts with the transcript.

Today, those transcripts are locked up in proprietary systems.

You have to pay a ton of money to get them out, or pay the hyperscalers a ton of money, or deploy a third-party solution, which also costs a lot and requires engineering effort to create an ecosystem to drive your downstream AI strategy.

The first product we’re launching today is Prism.

It is the first to market that is built on an open source model with a whole lot of pre-processing and post-processing know-how wrapped into it.

It can connect to your contact center environment in as short as 15 minutes and start showing you insights.

No code. No developing. No engineering. Nothing.

It’s priced per minute, per talk hour.

We’re at 90% cost reduction over most of the models and transcription engines in the market today.

What that does is it sets the foundation.

Now you own your transcripts. You own your call detail data.

That serves as the library.

It serves as your context.

It serves as compliance.

You now break free from having your data, your call recordings, your transcripts locked up in whatever proprietary system you’re paying a lot for every month.

We’re making it cheap, fast, easy.

This is the foundation.

Over the next 12 weeks, as we release new functionalities and features, you’ll be set.

If you already have transcripts today—awesome.

Let’s bring them aboard.

We can serve as your archival metadata layer now and forever.

I came up with this several years ago when I was traveling relentlessly and found myself flying Southwest a lot—not because they were always the cheapest carrier, but because they Just Made It Easy. I could book last minute. I could cancel last minute. I could change a flight last minute without fees, without calling. And so I found myself a very loyal customer because of that.

I started adopting that mindset into our companies—just make it easy for someone to do business with you.

Just this last week, I've reached out to a number of startups that I thought could help us, particularly with emerging AI capabilities that are interesting.

Only two of them actually made it easy to get on the phone, walk through a demo, walk through the product, and then actually made the buying process easy—very clear with their expectations and what they did and didn’t do. And, you know, they fumbled through the PowerPoints or whatever, which is fine. I still do that.

The rest of them made it almost impossible to do business with. I had one dude from a pretty well-funded startup where the soonest appointment they had was five days out. Then he was five minutes late for the demo, so I had already bounced, and he emailed me wondering why.

Just yesterday, I had a sales meeting booked, and it was declined by the founder because he made some assumptions about us as a portfolio instead of actually just taking a 15-minute call and realizing that I probably would have done two sales with him.

And those are just very recent examples. The list is long with folks who, for whatever reasons, are just making it difficult and introduce unnecessary friction just to do business with them.

The point here is to step back and look at all of your processes—whether it’s sales, onboarding, post-sales support—just make it easy to do business with, just make it easy to transact.

And man, you're going to accelerate a lot faster than being a pain in people's ass.

Subscribe now

There was a comment made to that post, which was quite insightful—part of the reason we still see so many companies doing the same thing is because of a lack of M&A activity in the market.

And I think that's quite astute because one of the most important pieces of advice I give a startup or a founder is to know the sea you're swimming in and to know it cold so that you can predict the currents before they change.

Every startup at some point is going to have a liquidity event.

You're either going to run out of cash and die, or you're going to go IPO—which very few companies actually achieve.

Or you're going to exit through acquisition, which is the most common way to realize liquidity, aside from going defunct.

But right now, very few companies are buying.

Rewind five years, ten years, fifteen years, twenty years—the M&A activity, particularly in contact centers, has been quite rich.

You've got NICE and Verint, both roll-ups built over two decades that started out as call recording and then diversified.

Beyond that, back in the day, you had Cisco, Avaya, and Genesys, all of them quite acquisitive.

At some level, Zoom was in the early days as well.

Fast forward to today, and really, in this post-GPT moment, the only companies getting acquired right now are the ones adding material cash or cash flow to the acquirer.

Take NICE—they’re only buying companies that will improve their quarterly earnings.

Look back two years—prove me wrong.

Five9 is interesting. I think they probably overpaid for Aceyus, but that was more of a tuck-in strategy that would have added some level of cash contribution to their earnings.

Zoom hasn’t acquired anyone since Solvvy a few years ago, probably because their organic growth is at a level where they don’t have to make a lot of acquisitions.

The FANGs—the Facebooks, Apples, Netflixes, and Googles—they’re also not buying right now.

They’re either building this stuff in-house or waiting for the market to mature enough.

The point of this is that no matter where you're at in the adoption or growth lifecycle, you need to be thinking about who the acquirers in your space are.

How do you align with them? How do you understand what’s important to them?

And then, at some level, how do you think of your company not as building products, but as the product—so that at some point, you’ll have that liquidity event or some type of acquisition.

For the last six months or so, I've been looking at AI in the context of the market, and it has felt pretty bubblicious—like we were hitting a bubble with a ton of companies jumping in. Some of them couldn’t even spell "contact center" a year ago. And some can’t spell "AI" most days. At this point, there are probably hundreds of companies doing pretty much the same thing.

Most of them are either focused on voice automation and chatbots—much easier to build today than even a year ago—or they're working on some flavor of call summarization, quality assurance automation, sentiment analysis, or CSAT. They take a call transcript, evaluate it against criteria through a prompt, generate output data, and put a front end on it—suddenly, it's a product.

We had already seen some of the air coming out of that, especially with the recent struggles of high-profile startups and the noise around deciphering real differentiation.

Then last week, DeepSeek released its R1 model, catching the market by surprise. It wasn’t a black swan event like GPT’s original release, which was a true paradigm shift, but it reset the market in many ways by unlocking new capabilities that didn’t exist a week ago.

Now, it's possible to run one of these high-performance models on commodity hardware, and cloud hosting providers are offering the full model at a fraction of previous costs. This shift opens up new possibilities for tackling complexity.

I no longer think the market was in a bubble. Instead, it had plateaued—everyone reached the same level, running in circles trying to figure out the next differentiation.

Some companies will recognize the opportunities that these new models and paradigms unlock, using them to create new products and categories. Others won’t, simply because these models introduce new capabilities that many don’t have the context to apply.

If a model touts PhD-level benchmarks, but you don’t have the expertise or the right questions to ask, you won’t get significantly different results from R1 than you would from GPT-4 or any other model. But those who understand the potential of models that don’t just generate code but can tell you what code to generate—or that don’t just create data but bring in vertical context to analyze it and surface new trends—will create an entirely new category of solutions.

We're going to see a divide: companies stuck at the plateau, having reached the limits of their competence and vision because they’ve only focused on automating what we already know how to do, and those who keep climbing, separating themselves by bringing entirely new capabilities to market.

Some of this will be driven by open-source models, making it possible to buy a Mac Mini and generate high-quality data 24/7 while leveraging more powerful models for reasoning-intensive tasks. The accessibility of these capabilities is shifting rapidly, and I expect we’ll see a wave of innovation until we hit the next summit.

That includes everything from: Will you be in business tomorrow? If I'm going to write you a check and build parts of my organization around you, how can I trust that you'll have the financial resources and ongoing ability to deliver?

A recent example is this week when the whole DeepSeek controversy hit the internet. The open-source AI model came out from China, and there was a lot of fear and uncertainty in the market about what that meant for people. If it was deployed, would data be going overseas to China? Could it cause some type of security issue?

Then there was the camp we’re in, which embraced this new technology pragmatically to ensure the success and integrity of our business and our customers.

Through all of that, not a single customer reached out to us with concerns about what we might be doing with their data and this new DeepSeek model. I believe the reason for that is trust.

One, contractually, we have a whole lot of safeguards in place to ensure data integrity—making sure data doesn’t leave the United States, strict access levels, HIPAA controls, and SOC 2 compliance. For us to start moving data around would be a breach of trust. It would be a breach of contract.

But beyond that, our customers also know that we have the vision and pragmatism to be good stewards of their interests. That’s what building a reputation for integrity and trust in the industry comes down to—demonstrating that you are a good steward.

I see us as stewards of our customers' data and their money. They invest with us every day, and we are responsible for delivering outcomes and results.

As a startup, you have to demonstrate that level of maturity and leadership day in and day out—whether that’s through contractual agreements that create safeguards for clients if something goes wrong, or, more importantly, by proving every day that you’ve earned their respect.

You need to show that you understand the market and these issues as well as anyone, and that you always have your customers' best interests in mind.

While the world is melting down over what’s the right answer or the wrong answer, sometimes the best thing you can do is reach out to your trusted advisors, get educated, and then relay that back to your customers so you can lead them through the process.

As you do that, you not only build trust and engagement, but you also build long-term relationships. That increases customer lifetime value because they know you're not going anywhere—and they know that you have their best interests in mind.

There's a lot of, I think, misguided but well-intended advice that the very first thing you need to do is figure out knowledge management or a knowledge base.

And while that is a critical component for a lot of reasons, it's not the very first thing you need to do. The first thing you need to do is ensure that the underlying technology you're acquiring—meaning your CCaaS, WFM, monolithic QA, or whatever applications or agent assistants you're implementing in your contact center—is open and extensible.

I talk a lot about this: APIs and the ability to move data and knowledge back and forth between those applications.

What will inevitably happen is that each of these use cases or applications in the contact center—whether it's an agent assistant, voice AI, or anything else—will have its own onboard knowledge management. Each one handles knowledge differently with memory and vector embeddings.

So, you need the ability to move your knowledge in and out of these systems in an open and easily facilitated way. This means you don't need a knowledge base; you need knowledge orchestration between these applications.

That starts with the ability to push out knowledge.

As an example, we deployed Zoom's agent assistant. It has its own knowledge base, which is very utilitarian in nature. The only thing it does is power Agent Assist, but it has to work that way because of how it indexes all the knowledge.

We have our own knowledge management platform that is open and extensible. You can create all your knowledge base articles in one place and, with just a few clicks, synchronize them directly into a Zoom knowledge base.

If we didn't have this synchronization capability, you'd have to administer knowledge separately in Zoom, in your IVA, and elsewhere. By the way, we synchronize everything out and back in.

The real point here is that if you're just going out to get a new knowledge management system, you're likely adding another tool in isolation. And if you're procuring and selecting your CCaaS, agent assistants, or any other system that doesn't have open APIs, you're just going to keep proliferating these applications without realizing their full value.

So, coming back to how we've designed this and how we guide our clients: it's about thinking through knowledge orchestration with document management and knowledge articles.

As we select partners to deploy voice AI, agent assistants, or even quality assurance, we can now perform deep quality assurance management. For example, we can embed a script directly inside a QA form or assess agent product knowledge right inside an evaluation using an LLM.

The only way to achieve this is by ensuring those vendors have published API documentation and that you can access and use it.

If you need help with that, let me know.

The current conversational AI vendors are doing quite well and are maintaining healthy profit margins. I've heard one mention an 80% profit margin number before. I'm just curious about your perspective on 10 to 15 years from now. When it comes to the ability of conversational AI vendors to maintain profit margins in that ballpark,

I don’t think they will. I think 80% gross margins are achievable because the technology is becoming so cheap so fast.

We were in conversational AI for a while, and back when we were doing it, it was much more expensive and complex. Now it's gotten to the point where there are so many open-source tools, and most of the intelligence is offloaded to the LLMs. You still have to deal with the plumbing, like voice connectivity, but even that is becoming super easy. What’s happening is a big proliferation of startups saying, “Hey, we’re another voice automation company, another conversational AI vendor, or another bot.” The market is saturated, which leads to compression of margins.

In 18 to 24 months, you're going to see CCaaS vendors incorporating this technology into their platforms. I know for a fact that at least three of them are working on it. They haven’t released anything yet due to the risks of hallucinations and wanting to ensure they get it right. Pricing is also a factor since they’re used to charging per seat. Now they might need to adopt pricing models like per minute or per transaction. And, as always, price approaches marginal cost in the long run. Bezos’s famous saying, “Your margin is my opportunity,” applies here. Once something is no longer complex to do, it’s time to move on. Voice automation will likely remain big and noisy for another year, but the momentum is waning because it’s no longer difficult or complex.

When I talk to contact center leaders, the gap between interest in voice AI and its adoption is striking. Everyone’s curious and eager to deploy something eventually, but they’re cautious because of hallucination risks. It’s interesting that CCaaS platforms are developing their own tools. This could align with a point where the technology advances enough to be confidently used with low hallucination risks, just as CCaaS vendors make it available natively.

This ties into the earlier discussion about the delay between interest in a product and its adoption. We’re in the early stages of the technology diffusion curve, with larger brands waiting to see what happens. If you’re a startup in that waiting period, you’ll likely be cash-starved. Early wins might come from digitally native companies, but not from large banks or laggards, which take longer to adopt. There’s a market for niche, vertical applications, like a healthcare benefits bot connected to marketplace APIs with specialization that narrows hallucination risks.

Good general startup advice: if you’re managing complexity on behalf of a customer, they’ll pay you. For example, integrating with systems like Epic in healthcare is a pain. Solving that is less about the voice AI and more about connecting into the infrastructure to facilitate transactions. That’s what you’re doing with ProxyLink—facilitating the transaction effectively.

When I see generic voice bots, I’m not excited. It feels like a CCaaS play or a niche-focused play.

Subscribe now

My response dove into the complexities of market timing, lessons from past ventures, and how redefining our go-to-market strategy turned challenges into a scalable platform and successful exit.

Subscribe now

One of the hardest things—and probably not really talked about a whole lot around being a founder—is timing and gauging where the market is and where the appetite is. Then, tapping into that early traction while you build toward the bigger vision.

What this sounds like to me—and I’ll tell you about Experian, which I think is super relevant here—is that the only two times I’ve really lost money in a venture were when I went direct to consumer. Your total acquisition cost or customer acquisition cost compared to customer lifetime value is typically inverted. It’s crazy expensive to acquire a customer because you’re competing for very limited wallet share.

I mean, your $10 a month or $20 a month product is competing against everything else. So, the customer has to have a problem right then in order to pull out the credit card. And then you’ve got to keep them as a customer.

If you look back 10 to 15 years at LifeLock, the identity monitoring company, they were pretending to be a tech company, but they were really a marketing company. They would spend $70 million a year—or per quarter—just on marketing to acquire customers. They also had a big retention problem.

When we launched our second company, we went after social network monitoring for teenagers and kids—for cyberbullying, sexual predation, and reputation management. This was back in the early days of social media—around 2012, I think, when we launched.

We realized that you’re asking your customers and prospects to adapt their value system in order to purchase your software. What happened was we were competing against a couple of other companies. One of them sold—I forget who they sold to—but their COO and I met one day. He said, “Man, you’ve just got to abandon this company. Walk away from it. There’s no business here.”

But what we did—and I don’t call it a pivot as much as just redefining our go-to-market strategy—was I went to the identity monitoring companies and said, “Look, what we’re doing isn’t a standalone product, but it’s a great feature in your product, in your bundle.”

So, what we ended up doing was becoming a platform—a back-end platform. We built out all the APIs so we had all the integrations into the social networks. We had all the algorithms to monitor for stuff. Then we went to these companies and said, “Just embed us. We’ll make it super cheap because our cost is so low.”

And they all did—all of them except LifeLock, which is funny.

We ended up selling that company to Experian because Experian has one of the largest consumer bases for identity theft monitoring. They’re a credit union, but they’ve got all these consumer products you’re talking about, right? So, they ended up rolling us into that.

What that did was take our core product and our mission—which really stayed the same. We wanted to get kid protection, call it, at scale.

DoNotPay, by the way, has raised $26 million, and I suspect a great chunk of that isn’t going into the tech—it’s going into what we’re talking about with acquisition costs. But we found that distribution, right?

So, we focused on the infrastructure, the algorithms, and all of that because it is kind of a two-sided market in some aspects.

Recently—this week alone, in fact—I’ve had three abysmal experiences where we were ready to buy something from a company, and they completely botched it.

Take Brex, for example. It’s a fintech and banking company. We did a lot of research to find banking services that could meet our needs across multiple entities. We were excited to get started. But then, they involved three different people across support, some startup consultant/advisor thing, and sales. None of them coordinated. All we needed to do was upgrade our account to support more than two entities, but nobody could figure that out. Instead, we got dragged through their script of pre-qualifying questions that had absolutely nothing to do with our business. So, short answer: we’re taking that business somewhere else.

The same goes for online support. We’re trying to embed support into our platform to create an elegant, world-class customer experience. But the company we approached started reading off their list of questions about who we are and how we do things today—which isn’t what I want. I want them to focus on who we’re going to become tomorrow and the level of support we aspire to deliver. When you ask the typical SDR or BDR to reframe their questions or guide them toward how you want to buy, it seems to short-circuit their script. Short answer: I’m not sure we’ll go with that company either.

Another company—this time for payroll—couldn’t even manage to respond about how to get a demo.

The moral of the story? Sit down and figure this out. Your onboarding and initial experience are your first impression with a client. If you can’t get that right, you’ll lose them. Retention will be horrible. Ironically, one of these companies is out there flexing about being in “founder mode.” My advice? Just focus on executing and delivering the type of customer experience you’d want yourself. Do that, and people will pay you—and stick around.

Subscribe now

I'm going to step back and look at it a little differently. First, I'm not distinguishing the difference between a BPO contact center and any other contact center that's in-house, because at the end of the day, we're all trying to accomplish the same thing: delivering compelling and excellent customer experience at the best price point possible. AI or no AI—that’s the end goal.

In BPOs, it's our job to lead the industry in how that gets done.

What I think about every day, when I sit down with my morning or afternoon coffee, is: How do we be the best at that? How do we deliver the best customer experience in the best-run contact center? That's the goal for BPOs on the planet. And that leads to an interesting set of questions. Notice I didn't ask, "How do we survive the AI apocalypse?" or "How do we deploy AI?" or anything like that.

The answer starts with just basic principles in economics and running and growing a business—or a service business, wherever you are.

The first operating principle is that we need to increase our capabilities—what we can bring to our customers and to the market, whether it's through products, services, or better execution of what we already do. We need to increase our capacity to deliver those things, and we need to do it in a way that consumes fewer resources. If you can bring those three things together, you're going to have a comparative advantage in the market.

Last year, we grew faster than we ever have. What we did was look at all of our constraints across the organization and ask: If we landed 1,000 seats, 10,000 seats, or even 100 seats today, where would things break down? Or where are things already breaking down with the business we have now?

Then, how do we leverage new tools and new ways of thinking with AI to solve for those and unlock new potentials? We started with training and recruiting, which is a pain point for every contact center on the planet. Attrition is a problem, no matter how good you are, how great your culture is, or how great your pay is.

You're constantly hiring. If you're growing, leaning into clients, you're hiring and training. And, frankly, our training program was a mess. We were throwing people at it, but finally, we stepped back, refactored the whole thing, and asked: How can we use AI to help us scale, improve the efficacy and outcomes of training, and then scale it so that we can do what I call "scaling in place"?

That means: How do we go from 1, 10, 100 seats, or 1,000 seats of new business without adding incremental headcount—except, really, at the agent level?

So we rebuilt the whole program from the ground up. What we have today is a set of great training modules and world-class content. Now, clients are coming to us asking, "Can you do the same thing for us? Can you build these modules or license yours to us?" That's created a new revenue stream for us and might even create a new business.

With recruiting, we've done the exact same thing. Again, recruiting is something every call center in the world has to do, and ours was painful. I strongly believe that if you can improve your talent acquisition strategy, you're going to improve retention, quality, and everything downstream.

We took a very data-driven and AI-driven approach. We've shrunk the time it takes to fill a new hire class, improved the quality of people in those classes, and improved the retention and quality coming out of that.

We're focusing on everything else before worrying about AI agent assists or any of that. We've unlocked new revenue, increased our capacity, and scaled as we add new logos and headcount.

The next piece is selecting and deploying technologies in production. Every CCaaS player has their AI. Tom Weird calls them "Whatchamacallit bots"—automation tools, voice automation, etc.

Not all BPOs are created equal. The industry isn’t monolithic or homogeneous. If you've rented thousands of seats offshore for simple phone calls, like minimum-wage workers checking order statuses, that’s a high-risk profile in the industry. We're not in that business.

For everyone else, it's about delivering these new capabilities better than yesterday without blowing your margins.

That's why we're big fans of Zoom. Their API documents let us build and deploy whatever we need, and their pricing is clear. They ship new code monthly, so we're constantly getting more for our dollar—and so are our customers.

If I look at Nice, Five9, Genesis, or Amazon Connect—most of the BPO world’s go-to tools—you're paying through the nose. Nice is like $250 per agent per month. Amazon Connect is in the same range. You're adding $1-$1.50 per agent hour just to get out of bed in the morning. No thanks.

With Zoom, we’re running a good piece of our business on their Essential licenses for $65/month. We get transcription, APIs, and analytics. Our analytics make the big kids cry themselves to sleep because we bring that to market.

When we move up the stack, we run on Elite licenses for higher-complexity calls. We wanted agent assist, better accuracy, workforce management, call quality, and screen capture—all in one.

This is how we show up fully loaded to any deal or opportunity. We're delivering capabilities, capacity, and results at a price point nobody else can match.

To me, that's how we're growing over the next 12 months—and forever. It’s an operating model: VXOS. Swap out Zoom or any tool, and you still follow the same principles.

AI is just a tool to eliminate constraints, unlock new capabilities, and scale. These are things everyone should be doing, whether you’re 5 seats or 50,000.

Our role in the industry is to help lead, deliver these capabilities to the market, and share what works. We've got scars from failed experiments—some we built and had to rebuild—but when we find things that work, we hit on all cylinders and scale.

That's our opportunity—even to partner with other BPOs. Some very large BPOs understand what we're doing with our sister company, Endeavor CX, around analytics and operational excellence.

Ultimately, if you’re competing for business, you're being measured on call quality, occupancy, and comparisons to everyone else. Your job is to become the best version of yourself. Stop worrying about competitors. Unlock your own potential and drive up your CX delivery expertise.

Subscribe now