By James Proctor, Co-Founder and Managing Director, The Inteq Group
Time to value from AI is largely decided before any technology is configured, at the moment the first process is selected. Choose a process with measurable value at stake, contained scope, identifiable decisions, and data that already exists, and value arrives in a quarter. Choose a sprawling process, or open with an architecture program, and value arrives after patience expires, if at all. Acceleration, in other words, is not primarily an engineering achievement. It is a selection and sequencing discipline, and organizations that treat it that way stop being surprised by their own timelines.
What Makes a Process the Right First Target?
Four criteria, applied ruthlessly. Real money, measurably at stake: cycle time, win rate, penalties, working capital, something a CFO already tracks. Contained scope: one process, one owner, one team feeling the pain, because political surface area is schedule risk. Identifiable decisions: the process's delays should trace to a small number of nameable decision points, not to diffuse dysfunction. And existing data: the decisions must be supportable, even imperfectly, by information the organization already holds. Note what is not on the list: strategic visibility, executive fascination, and vendor enthusiasm, three selection criteria that routinely produce eighteen-month science projects.
What Should the First Ninety Days Produce?
A baseline, a redesigned process, and a measured delta, in that order. Baseline first, because value nobody can verify is value that does not politically exist: instrument the current state, quote turnaround, aging, error rates, before touching anything. Then the decision-centric redesign of the one selected process, running in production conditions, not a sandbox. Then the published comparison. Insurance brokerage submission intake shows the shape. Commercial submissions arrive as unstructured broker emails and inconsistent ACORD forms, and speed to a competitive quote is a primary determinant of win rate, which makes the value line brutally simple. An intake redesign, agents assembling and structuring submission data, routing clean submissions to rating immediately, surfacing incomplete ones for targeted completion, moves quote turnaround within weeks, and the producers feel it in won business they can name. Nobody argues with a won account.
Guard the ninety days against scope accretion, because success attracts it. The moment the intake work shows promise, adjacent teams arrive with reasonable requests: add renewals, add another line of business, integrate with the portal project. Every addition is individually sensible and collectively fatal to the timeline. Park them, visibly, on a sequenced backlog. The discipline is not stinginess. It is the recognition that the first initiative has exactly one job, producing undeniable proof on schedule, and everything that dilutes that job is deferred value, not added value.
What Does the First Win Actually Buy?
The second one, and on better terms. A program's first measured result converts the conversation from whether to where next, and it recruits the scarcest resource in any transformation: operating leaders volunteering their processes. Treat the portfolio as a pipeline of quarterly proofs rather than a moonshot with a distant payoff, and each installment funds the next politically as well as financially. Acceleration, sustained, is just this compounding done deliberately.
“Your first deliverable is not a platform. It is proof.”
What Kills Acceleration?
The architecture-first instinct, and I will put the point at its full strength: if your AI program's first-year deliverable is architecture, you have already lost. The market for executive patience with AI has repriced, the deference that funded two-year foundations in 2023 is gone, and a roadmap whose opening act is infrastructure is a resignation letter written in advance. This is not an argument against architecture. It is an argument about sequence. Proof funds architecture. Architecture has never once funded proof, and every quarter spent building capability nobody can see is a quarter of political capital burning at compound rates.
Selecting the first process against the four criteria, and designing the baseline-and-evidence plan around it, is exactly how our agentic AI consulting engagements typically open, because the selection meeting is worth more than any technology decision that follows it.
