The 100-day AI plan after close

Post-close bandwidth is scarce. Without a written AI workstream, companies either freeze (and lose a year) or thrash (and ship risk). The 100-day plan is how sponsors convert diligence findings and investment-thesis AI claims into an operating agenda. It is deliberately boring in places — inventories, contracts, budgets — because boring foundations are what let AI value creation compound later.

Use this alongside the AI in private equity playbook and the checklist in the operating library. If diligence produced an issues list, that list is an input, not a souvenir — see AI due diligence.

Inventory every AI tool, API key, and shadow chatbot. Pull invoices for model providers and relevant cloud lines. Identify who has admin access. If keys are scattered in personal accounts, rotate them into a governed path — even a lightweight one — before usage explodes. Capture data-rights and security findings that block customer-facing work.

Appoint names, not committees: executive sponsor (CEO or COO), technical owner (CTO/VP Eng), value-creation liaison from the fund, and a security/privacy contact. Schedule a recurring weekly standup for the AI workstream. Announce a temporary policy: no new production AI features without sponsor visibility.

Rank use cases with the scoring logic from the value-creation spoke: volume, pain, readiness, change difficulty, and unit economics. Select two or three. For each, write a one-pager: baseline metric, target, eval method, data sources, model routing hypothesis, budget cap, and kill criteria. Wire metering tags even if the shared gateway is immature — you cannot manage what you do not tag.

Estimate inference with the inference calculator and rates from the pricing index. Decide where frontier vs efficient models apply using open-weight vs frontier. If governance gaps are material, schedule remediation tasks in parallel and keep pilots internal until cleared — see AI governance.

Launch pilots to limited user groups. Collect qualitative friction and quantitative metrics weekly. Watch for cost anomalies and quality regressions. Escalate blockers within 48 hours — waiting until day 90 to admit a data-access problem wastes the plan.

At the end of this window, graduate, iterate, or kill. Killing is success when evidence is clear; zombie pilots are how credibility dies. Document learnings in a form the next portco can reuse via portfolio AI operations.

Produce a board pack section: spend, value metrics, risks, decisions needed. Refresh the ranked backlog for the next two quarters. Formalize RACI for incidents and change management. Hand steady-state dashboards to the owners who will live with them after the value-creation surge team steps back.

Tie inference FinOps rituals into the monthly CFO cadence — details in inference costs. If the company will join a firm gateway or lab vehicle, schedule the technical onboarding rather than leaving it as a slogan.

Step 1: Inventory tools, keys, vendors, and spend. Step 2: Appoint owners and freeze unmanaged production keys. Step 3: Close or schedule critical governance remediations. Step 4: Select two or three pilots with baselines and kill criteria. Step 5: Implement metering and budget caps. Step 6: Build eval golden sets. Step 7: Launch limited pilots. Step 8: Review weekly; kill or iterate. Step 9: Write the board narrative. Step 10: Lock the next-quarter backlog and platform handoff.

Customize heavily for regulated industries and for turnarounds where AI is not the binding constraint. The plan is a scaffold, not a religion. Market context on how peers staff deployment remains on the Deal Wire.

Weeks 1–2. Kickoff, inventory, access freeze for unmanaged keys, spend pull, diligence issues import, RACI published. Weeks 3–4. Workload workshops with top functions, draft scoring of use cases, legal review of top vendor terms, interim budget caps. Weeks 5–6. Pilot one-pagers approved, baselines measured, eval sets v0, metering tags live. Weeks 7–9. Limited pilot launch, weekly metrics, rapid blocker removal. Weeks 10–12. Kill/iterate/ graduate decisions, board draft, backlog lock, platform handoff checklist.

Adjust pacing for carve-outs, regulated sectors, and crisis turnarounds. If the company cannot close books reliably, AI copilots are not the binding constraint — say so explicitly and reschedule rather than running a Potemkin 100-day AI theatre.

Keep a visible RAID log (risks, assumptions, issues, dependencies). AI workstreams die from hidden dependencies: data access, SSO delays, vendor DPAs, and exhausted engineering bandwidth during peak season.

By day 100 you should be able to hand a successor: (1) inventory and spend baseline, (2) routing and quality- floor policy v1, (3) pilot scorecards with evidence, (4) risk register with owners, (5) board section, (6) next-quarter backlog with estimates, and (7) a decision log explaining kills. If any of those are missing, the plan is incomplete even if demos impressed.

Store artefacts in a shared drive with access control, not in personal folders. Future you — or a future operating partner — will need them during refinancing, add-on diligence, or exit. The operating library templates are starting points; customize ruthlessly.

Celebrate clarity. A clean kill of a bad pilot is a cultural win worth announcing. It teaches the company that evidence matters more than AI fashion — exactly the culture that produces real value creation later.

AI is one workstream among pricing, working capital, talent, and integration. Compete for CEO attention honestly. Bundle AI tasks into existing forums where possible rather than inventing a parallel universe of steering committees. If a pricing initiative and an AI initiative both need the same data warehouse team, sequence them.

Deal teams should attend the kickoff with their diligence evidence packs. Operating partners should bring portfolio standards. Portco leaders should bring the political map of what will actually get adopted. When those three seats are not in the room, plans drift into abstraction.

At day 100, schedule the day-180 review now. Continuity is the difference between a burst of activity and a hold-period capability. Tie that review to FinOps and governance cadences so the workstream has a home after the surge.

Across funds, the teams that make durable progress share a few habits. They write decisions down with dates. They refuse to expand scope before metering exists. They pair every automation claim with a quality floor and a named executive owner. They bring CFOs into model-routing debates early, before unit costs become a surprise in the monthly pack. And they treat vendor press releases as inputs to diligence, not as substitutes for operating proof.

The teams that struggle also rhyme. They launch too many pilots. They staff AI as a side project for already overloaded engineering managers. They buy enterprise agreements to “get started” without workload maps. They hide failures instead of killing them. In a five-year hold, those habits compound into wasted calendar time — the scarcest resource in a portfolio company fighting day-to-day fires.

On AI100DayPlan, use the rest of this site as a toolkit, not as dogma. The Deal Wire tells you where capital is forming. The league table shows who is participating. The pricing index and calculator quantify unit economics. The spoke guides dig into sourcing, diligence, costs, value creation, ops, governance, model choice, and the first hundred days. Your job is to assemble the pieces into a plan your board can govern and your operators can run on a Monday morning.

Finally, remember the asset-class basics still bind. Returns still come from buying well, improving companies, and selling better. IRR and MOIC still disagree usefully. Leverage still amplifies both directions. AI changes the operating toolkit and the cost stack inside that timeless loop. If you keep that proportion straight, you will ask better questions than peers who think a model alone is a strategy.

Practitioners should leave this page with a bias toward instrumentation and accountability. Write the metric before the pilot. Write the owner before the vendor. Write the kill criteria before the kickoff. In private equity, calendar time during the hold period is the inventory you cannot replenish — spending it on unmeasured AI activity is still a real cost even when the invoice looks small.

Share learning across the portfolio ruthlessly. A failure documented in one company is a gift to the next. A success that remains tribal knowledge in a single CTO’s head is an undiversified asset. Sponsors that build that learning loop — alongside capital structures they already understand — will treat AI as what it is becoming: a standard chapter in value creation and risk management, not a side demo for visiting LPs.

Continue through related guides linked on this page, keep as-of dates on every figure you reuse, and return to primary sources when a Deal Wire entry matters to a live decision. Good process compounds quietly; that is usually what good returns look like from the inside.