When a company commits to AI in public

A company announces AI agents in customer service, or an AI-first strategy, or a rollout across finance. The commitment is public. The team to deliver it safely usually isn't in place yet.

That gap is the signal.

Why the announcement creates work

Executives announce AI programs for real reasons. Investors ask about it on every call. Boards ask about it at every meeting. Competitors are announcing their own. Saying something in public is often how a CEO gets the organization moving.

But the announcement runs ahead of the capability. The data isn't clean enough. Nobody has decided who approves a model before it touches customers. Security hasn't reviewed what the agent can access. The people whose work is changing haven't been told how.

Once the CEO has said it out loud, someone inside has to make it true, on a timeline that was set for the press release. That person needs help with data, integration, evaluation, governance and change management. Very little of that is the model itself.

Where the commitments show up

Earnings calls are the main source. Listen for a named function and a timeline. "We're deploying AI agents across our claims operation this year" is a program. "We're excited about the potential of AI" is a sentence every CEO says.

Press releases announce partnerships with model providers and platform vendors. Those tell you the technology choice is made and the delivery work is next.

10-Ks increasingly carry AI language in the business description and the risk factors. A new risk factor about AI governance, data privacy in AI systems or reliance on third-party models is leadership admitting they haven't solved it.

Job posts tell you what's actually missing. A company that announced an AI strategy and is now hiring for an AI governance lead, ML engineers and a head of AI product is building the team after the promise. And a new chief data or AI officer is often the clearest tell of all, which I covered in When a CDO arrives, data budgets follow.

Separating commitment from noise

Strong language names a function, a use case and a date. Customer service agents by Q3. AI-assisted underwriting across commercial lines this year. Those are programs with owners.

Strong language attaches a number. A cost target or a productivity goal tied to AI means finance is tracking it, and someone's bonus depends on it.

Weak language is aspirational. "Exploring," "investing in capabilities," "AI-powered future." That's a company talking, not building. Watch it, don't write to it yet.

And be honest about whether your firm fits. If you're an AI firm with real delivery history, the gap is obvious. If you're a data engineering or integration firm, your angle is the foundation the AI depends on. If you're a change management or security firm, you're in too. Just don't pretend to be an AI firm when you aren't. Buyers see through it fast.

Who owns the promise

The CEO made the announcement. Someone else owns delivering it. Usually that's a chief data or AI officer, a CIO, or a chief digital officer. In a function-specific rollout, the business leader for that function owns the outcome, the head of customer service or the CFO for a finance rollout.

The CISO and general counsel feel it from the other side. They're the ones who get blamed if an agent leaks data or says something it shouldn't. They're often underserved by sellers and very open to a conversation about doing this safely. See Selling to a CISO.

An opener that isn't another AI pitch

Every buyer who made an AI announcement now gets a flood of messages about AI. Almost all of them open with the vendor's capability. "We help companies accelerate AI adoption." Delete.

The way out is to stop talking about AI and talk about the specific thing they promised and the specific thing most likely to break.

Say a regional insurer announced on its earnings call that it would roll out AI agents across claims intake this year. Here's an opener to the head of claims.

Your earnings call committed to AI agents across claims intake this year. You're the one who has to make that work without slowing claims or creating a compliance problem. My guess is the agent is the easy part and the hard part is the claims data and who signs off on what the agent can decide on its own. Is that close, or is the harder problem somewhere else?

The opener never pitches AI. It names their promise, puts it on their seat, and guesses at the unglamorous problem underneath. That's what gets a reply. More on building that guess in Hypotheses for data and AI services.

Timing

The first quarter after an announcement is when leadership figures out what it actually takes. That's your window. By the second or third quarter, they'll either have a partner, have scaled back, or be in trouble.

Trouble is an opening too. A rollout that slipped, or a quiet walk-back on the next earnings call, means the first approach didn't work and someone needs a better one.

Stack the announcement with hiring, a new data or AI leader, or a new risk factor. Any two of those together tell you the program is real and the team isn't there yet.

A public AI commitment is a promise someone inside now has to keep. Write to that person about the hard part, and leave the hype to everyone else in their inbox.