By Damian Mathews & The Last Mile Team
I have a question for every CX leader running a careful, measured, watch-the-market-mature approach to AI right now.
What’s your bet?
Because that’s what it is. The decision to wait, to let the technology settle, to see how the dust falls before committing real resources, is a bet. And most of the leaders I talk to haven’t thought clearly about what they’re actually betting on.
There are two possibilities for where AI is heading. The first is a ‘bubble.’
The valuations are inflated, the use cases are thinner than the marketing suggests, and at some point capital flees, the hype collapses, and the technology takes its place as a useful but unremarkable layer in the stack. Like cloud, but louder. If this is what’s happening, the patient leader wins. You wait. The bubble pops. The smoke clears. You buy mature tools at reasonable prices and integrate them at a sensible pace. Conservatism is correct.
The second possibility is displacement.
The technology is real, the capabilities are escalating, and the organizations adopting now are not just getting ahead. They are restructuring the basic economics of their industries. The competitive position your contact center holds today depends on cost structures and operating models that AI is dissolving. If this is what’s happening, there is no correction to wait for. The gap between the organizations adopting and the ones watching just keeps widening, and one day you’ll wake up and realize you can’t catch up. Conservatism is no longer conservative, it’s suicide.
Two scenarios. Opposite responses. Many leaders I hear from are defaulting to scenario 1, without properly engaging in scenario 2. Are you one of them?
Fish wrote about this asymmetry recently in Heisenberg’s Peak. His argument is that the wave of AI capability hitting the economy right now isn’t a single wave that will crest and recede. It’s a cascade, with each new wave of real capability arriving before disillusionment with the previous one can settle in. If that pattern holds, the trough most leaders are waiting for never arrives. The correction never comes. And the organizations that were waiting for it just keep falling further behind.
That’s a strong claim. But you don’t have to take Fish’s word for it. Just look at the evidence sitting in front of every CX leader right now.
A plumbing software startup hit a billion-dollar valuation by deploying AI voice agents in weeks, while most enterprise contact centers are still in month nine of a vendor evaluation. Coinbase publicly restructured itself around AI as the operating core. PwC found that 20% of organizations are capturing 74% of all AI-driven economic value, and that gap is compounding, not closing.
None of those data points describe a bubble. They describe displacement happening in real time, in specific sectors, with no correction in sight.
Here’s the part that should give every CX leader pause. The asymmetry of the bet is not symmetric. If you’re wrong about the bubble, you save some money in the short term and miss the upside. If you’re wrong about displacement, you lose your seat at the table permanently.
That’s not hyperbole. Kodak invented the digital camera in 1975. They sat on it for two decades because they didn’t want to cannibalize their film business. They had the technology, the talent, the distribution, the brand. They chose to wait and see how digital photography played out. By the time they committed, the seat was gone. They filed for bankruptcy in 2012. The lesson, of course, isn’t that incumbents always lose. It’s that the cost of being wrong about a structural shift is categorically different from the cost of being wrong about a cycle.
Which means the right question for a CX leader today is… “What would change about our approach if we genuinely believed displacement was happening? Would we still feel comfortable waiting?”
If the answer is no, the work is to figure out what moving looks like in your environment, on your existing stack, with the governance and security your business demands.
That’s the system we’ve been building toward, and Kerry documented our own version of that journey in A1B: Customer Zero to AI-First.
What are you betting on?
— Damian
Early signals you should keep on your radar.
Nvidia signed a $3.4 billion, five-year GPU cloud deal with IREN and committed to deploy up to 5 gigawatts of DSX infrastructure at IREN sites. IREN handed Nvidia a five-year warrant on 30 million shares at $70, tighter than a typical supply contract. Tier-two operators locking in hyperscaler-scale capacity could leave less GPU oxygen for everyone behind them.
Cerebras priced its IPO above range, set to raise up to $4.8 billion at a $48.8 billion valuation with the book roughly 20 times oversubscribed. The wafer-scale chipmaker is betting the company on inference, anchored by a $10 billion OpenAI supply deal and a March AWS pact. Demand at this scale suggests investors believe inference is where the next chip margins may live.
Leading Edge
Proven moves you can copy today.
Anthropic launched Claude for Legal with twelve practice-area plugins and twenty-plus MCP connectors covering contracts, discovery, litigation, and M&A work. Freshfields, Quinn Emanuel, Holland & Knight, and Crosby Legal are already running it on live matters. Big Law going public with deployments at this scale puts pressure on every other firm and every legal tech incumbent to show their AI hand.
OpenAI spun up a $4 billion enterprise deployment subsidiary backed by TPG, Bain Capital, Capgemini, and McKinsey, then acquired AI consultancy Tomoro to staff it. Enterprise already accounts for over 40% of OpenAI revenue, with parity expected by year-end. OpenAI is moving from model vendor to systems integrator, which could squeeze the SIs partnering with it.
Off the Ledge
Hype and headaches we’re steering clear of.
Cloudflare laid off 1,100 staff, about a fifth of headcount, in the company’s first mass layoff, blaming AI for making the work obsolete. Internal AI usage jumped more than 600% in three months while Q1 revenue hit a record $639.8 million. Record-quarter layoffs are becoming the playbook; ‘AI made it obsolete’ may be the cleanest story whatever actually drove the cuts.
A new Lancet audit found a 12-fold surge in AI-fabricated citations, with nearly 3,000 of 2.5 million biomedical papers citing references that don’t exist. One in 277 PubMed-indexed papers in early 2026 contained at least one fabricated reference, up from one in 2,828 in 2023. Enterprise teams letting LLMs draft citations without verification may be quietly building the same trust debt now hitting academic research. Remember: YOU OWN THE OUTPUT!
