Skip to content
Fulkerson Advisors
Practice · No. 03

AI adoption that survives the front line

The hardest part of an AI rollout isn't the model; it's the team change. We work with Fortune 500 operators (litigators, store managers, buyers, clinicians) to make the new tool the path of least resistance. Adoption is the work, not the afterthought.

Discuss your rollout

Engagement model

What this looks like in practice

One of our three partners leads the engagement; you do not get handed to a junior bench. Bilal or Cynthia sits inside your operation for the duration, alongside an internal sponsor and an operating lead from your team. We work the floor: shadowing the people who will use the tool, reviewing the output they actually produce, and rewriting the prompts, rubrics, and integrations until the assistant earns a seat in the workflow. When we need depth (litigation process, retail labor cycles, clinical documentation) we pull from our expert network rather than improvise. Every artifact we leave behind (rubrics, training kits, feedback loops) is owned by your team, not licensed from us.

Buying trigger

When you should talk to us

The trigger is rarely "we want to talk about change management." It is usually one of three moments. A pilot scored well in the lab and stalled the moment it touched the front line; usage is flat and the steering committee is asking why. A board mandate has landed ("AI across the function by year-end") and the COO needs a credible plan for how 4,000 operators actually change their daily work. Or a tool is technically live, but the senior practitioners (the partners, the head buyers, the lead clinicians) quietly route around it, and the CFO is starting to ask what the spend bought. If any of those sound like the room you were in last week, we should talk.

Methodology

How we approach adoption

We treat adoption as a sequence, not a campaign. First we find the most credible skeptic on the team and earn their endorsement by tuning the tool to their judgment; nothing else moves until that anchor is set. Then we instrument the workflow so usage, quality, and business impact are visible weekly to both the operating team and the executive sponsor. Then we hand the cadence to your team, with a written rubric for where the assistant is trusted and where it is not. The sequence is boring on purpose. Most failed rollouts skipped a step.

  • Anchor with the most respected skeptic, not the most willing volunteer
  • Instrument three layers: usage, output quality, business impact
  • Codify a living rubric of trusted versus reviewed use cases
  • Run a weekly retraining loop on real production output
  • Hand the operating cadence to a named internal owner

Proof

Case files

The pattern shows up across industries. A top-5 US law firm needed senior litigators (not associates) to actually use an LLM agent inside pre-litigation intake; a large US retailer needed store managers to trust an optimization tool over their own gut on labor; a Caribbean conglomerate needed category buyers to abandon spreadsheets for a forecasting platform. In each case the model was the easy part. What follows is what the engagement actually looked like.

Anti-positioning

What we don't do

We don't run all-hands training roadshows; they generate slides and not behavior. We don't ship adoption dashboards that only report usage; usage without quality and business impact is a vanity metric, and senior leaders see through it within a quarter. We don't take on rollouts where the underlying tool is wrong for the workflow; if the model cannot do the job, no amount of change management will close the gap, and we will say so in the first two weeks. And we don't stay longer than we should. Our job is to make ourselves unnecessary.

Who you’ll work with

Bilal Bitar

Bilal Bitar

Co-Founder & Managing Partner

Bilal led McKinsey transformation engagements and structured operating-model change at the European Investment Bank, DEG, and PROPARCO; his INSEAD MBA work focused on how large institutions actually shift behavior, not how they announce it.

Practice background

  • Bilal Bitar has spent his career at the meeting point of finance, strategy, and large-scale transformation. He began in management consulting at McKinsey & Company before moving into investment banking, where he structured and raised development finance — channeling capital from European institutions including the European Investment Bank, DEG, PROPARCO, and Finnfund into energy infrastructure.
  • That grounding in capital markets and operational strategy anchors his advisory work today. He has led enterprise-wide transformation programs for organizations across banking, retail, and technology — aligning leadership teams and rebuilding the processes beneath them — and is fluent in the financial-modeling and value-creation frameworks that decide whether a transformation actually pays for itself.

Case files

  1. Top-5 US law firm

    Pre-litigation interview and document automation

    Senior litigators adopted an LLM agent for intake and document review after a three-month embedded rollout with the partner cohort.

  2. Large US retailer

    Retail labor planning and optimization

    Store managers ran the new optimization recommendations daily within six weeks; floor-level acceptance held above 80 percent at quarter end.

  3. Caribbean conglomerate, retail

    Demand forecasting Center of Excellence

    Buyers and category managers replaced spreadsheet workflows with the forecasting tool; the CoE has trained two cohorts of internal users since handoff.

See all case files
Frequently asked

Questions we hear most.

How do we get senior practitioners to trust an AI assistant?
Trust is earned in their workflow, not in a town hall. We embed with the most skeptical senior operator first, instrument the assistant against their judgment, and tune until the model's output is something they would have written themselves. Once one respected practitioner endorses it, the rest of the bench follows. We do not lead with mandates from above.
What does change management look like for an LLM in the loop?
It looks less like a training program and more like an apprenticeship. The assistant has to learn the firm's tacit standards (tone, risk posture, escalation rules) and the team has to learn where the assistant is reliable and where it is not. We codify both sides: a living rubric of where the model is trusted, and a feedback channel that retrains it weekly. That artifact becomes the operating manual.
How do we train the front line without grinding to a halt?
We do not pull people offline for week-long sessions. Training happens inside the work: short shadowing sessions, in-tool prompts, and a designated peer (often a respected mid-level operator) who fields questions in the first 60 days. Productivity dips for two to three weeks; we plan for it and protect the team's commitments during that window. By week eight, throughput is above baseline.
What KPIs prove the rollout worked beyond "adoption rate"?
Adoption rate is a vanity metric on its own. We track three layers: usage (daily active operators on the tool), quality (acceptance rate of the assistant's output without manual rewrite), and business impact (hours returned, cycle time, error rate, or revenue per operator depending on the use case). The third layer is what we present to the board.
Who owns the rollout: us, or your team?
Your team owns it; we run alongside. A named internal sponsor and an operating lead sit with our forward-deployed partners from week one. By the end of the engagement they are running the cadence, the rubric, and the retraining loop. We have no interest in becoming load-bearing.
What happens if a team refuses to adopt the tool?
We treat refusal as signal, not failure. Usually it surfaces a real problem: the model is wrong in a context we missed, the workflow integration is clumsy, or incentives are pointing the wrong direction. We diagnose which, fix it, and re-engage. If after that the team still declines, the honest answer is that the use case is wrong for that team, and we say so.

Bring us a question, and we’ll bring you an honest read.

Discuss your rolloutRead the case files