If it can see the final answer, use cleaned-up knowledge, and skip the pressure of the real workflow, it isn’t a pilot. It’s a demo. Pursuit Replay reconstructs a completed pursuit and replays it with AI under time-locked evidence and real controls — so you find out what would actually work before you risk a live opportunity.
Most AI assessments measure opinions, maturity scores, and model capability. None of that tests whether a redesigned workflow survives the pressure, ambiguity, knowledge gaps, and human behavior of a real pursuit. A replay does — because it’s run as an experiment, not a pitch.
At each replay step the AI sees only what your team knew at that moment. The final submission stays hidden — it’s the answer key, not an input.
Four measured conditions separate what the model contributed from what fixing your knowledge contributed from what your people contributed. Nothing gets attributed to “AI” by default.
Every workflow leaves with a verdict — retire, rework, shadow, copilot, or delegate — against thresholds agreed before the replay. The artifact is a flight certificate, not a vibe.
We reconstruct what actually happened on a completed pursuit — every action, wait, handoff, and judgment call as an event. The official process chart is fiction; the twin is not.
Your knowledge becomes an evidence graph: requirements connected to claims, claims to sources, sources to owners and approvals. Then we time-lock it — at each replay step, the AI sees only what your team knew at that moment.
Every workflow gets a one-page AI job description: permitted inputs, prohibited actions, escalation rules, and the human who stays accountable. Observe, recommend, or draft — external commitments stay human.
Four conditions, run as an evaluation: what actually happened, the model alone on your messy knowledge, the governed system, and your own team on the redesigned workflow. Held-out data, adversarial cases, blind review.
The deltas separate what the model contributed from what the knowledge work contributed from what your people contributed. The deliverable is an investment decision with conservative, expected, and upside cases — not a roadmap.
Starting ranges are published; the fee is fixed once the scope is. No win fees, on any contract type, ever.
$40K–$50K · 4 weeks
One completed pursuit, up to three replay workflows, a sandbox environment, and an executive readout with the production business case. Milestone billing: half at kickoff, a quarter at the twin, a quarter at the readout.
$60K–$90K · for regulated environments
The same replay inside your tenant or a virtual private environment: enhanced access controls, documented data flows, zero-retention where available, security and legal at the table, extended adversarial testing.
$100K–$250K+ · follow-on
Production implementation of what graduated: integrations, knowledge remediation, change management, workflow-owner enablement, and continuous evaluation. Each workflow ships with its own flight certificate.
scoped per cohort · workshops to embedded programs
Your people, trained on the workflows we install and on pursuit AI generally. Adoption is a measured delta in every Replay — training is how you move it. This is the spine of longer engagements, and it's fixed once scoped like everything else.
$12K–$25K/mo · 3–6 months
Portfolio leadership across your AI efforts after the replay: vendor guidance, measurement, coaching, and the judgment calls — without the executive hire.
Vendor-neutral means it: if the replay shows a subscription tool is enough, the readout says so. Inspect a full sample readout— all four artifacts, on a constructed pursuit. And everything we install is yours — the workflows, the evidence graph, the certificates. The Workbench is a live demo of the kind of tooling involved.
Retire what has no value or unacceptable risk. Rework what’s promising but not ready. Shadow what should run parallel to humans. Copilot what’s approved for supervised use. Delegate what earned narrow autonomy. Thresholds are agreed before the replay, so the verdict can’t be negotiated after.
Direct capacity value and speculative opportunity value are calculated separately and never combined into one headline figure. You get conservative, expected, and upside cases — the number a CFO can take to a board.
Four weeks later you’ll know what AI would have changed, what it would have broken, and exactly what to build — measured, not promised.
Book a Pursuit Replay →