A dinner in Midtown, and a very candid conversation about A.I.

Twelve executives from banking, insurance and a handful of other industries sat down to dinner. By the second course, they had stopped being polite about how ready they really are.
Image

THE NIGHT

The premise was simple: skip the panel discussion, skip the slides, and see what a room of people actually building with artificial intelligence will say to one another when nobody is being quoted for a keynote. What they said, over three courses of handmade pasta and a steady pour of wine, amounted to a case study in how far corporate America has come with A.I., and how little of that progress anyone can yet prove.

The setting supplied the ease that made people willing to discuss it honestly. The private room at Ramerino was quiet enough for conversation to carry across the table but not so quiet as to feel sealed off, with the low hum of the main dining room bleeding through the walls. The staff moved through the room with the kind of restraint that goes unnoticed when it is done well: plates cleared, glasses refilled, nothing rushed.

The food did its part. Three courses and dessert came out over the better part of two hours, generous enough to let the conversation wander without ever competing with it. By the later courses, a few guests had begun trading introductions on their own, the kind of quiet connections that turn into something more weeks later.

Everyone agreed on one thing before the entrées even landed: nobody is sitting this out. Pressed on how ready their organizations actually were, though, the honesty got sharper still.

“

We stopped asking whether to do this two years ago. The only question left is whether we’re doing it well.

BANKING EXECUTIVE
WHERE THE ROOM ACTUALLY STOOD
SURVEY SENT DAYS BEFORE  •  10 OF 12 RESPONDED
0
of the ten could name a hard return-on-investment figure
40%
had not settled on how they would measure one
30 / 30
percent named the CEO as owner; equally many said nobody
70%
see success as a blended human-and-agent workforce

THE RUNNING THEME

A room full of people moving quickly, with almost no one keeping score.

Nobody claimed to be standing still. But not one of the ten could name a hard return-on-investment figure, and 40 percent admitted they had not even settled on how they would measure one. Ownership proved just as unsettled: 30 percent named the chief executive, another 30 percent said in effect nobody, and not a single person pointed to a CFO, a COO or a business unit leader.

For a group that included several executives who described having already redesigned roles around the technology, the uncertainty over who is accountable for it landed as more than a punchline.


WHERE THEY ARE TODAY
40% researching and experimenting
30% running pilots
20% in production at scale

50% have already begun rebuilding roles around the mix; another 20% say they plan to.

70% hedge deliberately, buying vendor tools and building custom systems rather than committing fully to either.


“

We’re ready in pockets, which is a nice way of saying we’re not ready as a company.

EXECUTIVE AT THE TABLE
“

We’re building the plane while flying it. The trouble is, nobody’s checked the altimeter.

ATTENDEE, TO KNOWING LAUGHTER
“

Everyone wants to own the upside. Funny how few people want to own the rollout.

EXECUTIVE AT THE TABLE
“

We didn’t bolt A.I. onto old job descriptions. We rewrote them.

EXECUTIVE AT THE TABLE

 


 

WHAT IS HOLDING IT BACK

A perfect storm of process, talent and the plumbing underneath

Talent and technology tied as the most cited obstacles at 40 percent apiece, with data concerns taking the remainder. Governance and process control drew the sharpest exchange of the evening, not because anyone had solved it, but because the room saw how much of the work is invisible to the people accountable for it.

That points squarely at workflow modelling, work management and the human-and-agent control factors that sit between an ambition and a system you can run.

CITED AS BLOCKERS
Talent and skills40%
Technology and infrastructure40%
DataRest
“

How can you control that which you cannot see?

ATTENDEE, ON GOVERNANCE
“

The tech’s actually the easy part. Finding people who can run point on this, that’s where we’re stuck.

EXECUTIVE AT THE TABLE
“

We buy what’s commoditized and build what’s core. The mistake people make is not knowing which is which.

EXECUTIVE AT THE TABLE

WHAT STAYED WITH US

Less any single answer than the shape of the room itself: a group already deep into building with A.I., comfortable admitting in front of their peers that they still could not say, with confidence, whether it was working. That candor, more than any number from the evening, may be the truer snapshot of where corporate America stands on the technology it has spent two years racing to adopt.


WHERE THAT LEAVES YOU

The gap at that table was never ambition. It was a plan you could hold someone to.

Every organization in the room knew where it wanted to go. What none had was a named owner, a measurable target and a sequence for getting there. That is what a Vision Sprint produces: a short, structured engagement that turns agentic ambition into a plan your board can read and your operators can execute.

A scorecard, before you build
The ROI baseline and the two or three measures you will actually be judged on.
A named owner
Accountability that sits with a person and a function, not with the CEO by default.
A sequenced roadmap
Which work gets agentified first, which roles change, and what you leave alone for now.

Start with a Vision Sprint.

A few focused sessions with your leadership team, and you leave with the owner, the baseline and the roadmap that the room in New York was still missing. We are taking this format to other cities in the coming months, so tell us if you would like a seat at the next one.

Related content