Private equity and portfolios
100-day AI plan for a newly acquired portfolio company
By SourceX Editorial · Updated
Short answer
A 100-day AI plan for a newly acquired portfolio company should spend days 1-30 on a metadata-only records inventory and guardrails, days 31-60 on two use cases and the records behind them, and days 61-100 on decisions the board can fund. Every step has a named owner, and no records leave the company while the plan runs.
Key takeaways
- Start with a records inventory collected as metadata only: systems, years of history, record families and rights status.
- Put guardrails in place in the first month: an AI use policy, a review of vendor AI settings, and company-controlled admin accounts.
- Choose two use cases tied to records the company actually holds, rather than a long list of ideas.
- End the plan with board decisions for each option: fund, pilot, park, or assess the records for licensing.
- Keep vendor pilots from receiving company records until their terms are reviewed and approved.
What should a 100-day AI plan include?#
A 100-day AI plan should include four workstreams: a records inventory, guardrails, a small number of use cases, and a set of decisions for the board. It should leave out platform purchases, senior AI hires and long pilot lists until someone knows what records the company actually holds.
The plan works because each workstream feeds the next. The inventory shows which use cases are realistic; the guardrails stop records leaking while they are explored; the use cases give the board something concrete to fund or decline.
Starting with records reflects where mid-sized companies say AI stalls. In RSM's 2026 middle-market survey, data quality and availability issues were the most-cited inhibitor to AI deployment (34%), ahead of security and privacy concerns (30%). An inventory in month one surfaces those gaps before money is committed.
The plan at a glance#
The template below is a starting point for the operating partner and the portfolio CEO to adapt together. Replace each owner title with a person's name at the kickoff, and keep the plan to one page so it survives contact with the wider integration.
| Window | Workstream | Deliverable | Owner |
|---|---|---|---|
| Day 0 | Kickoff | Plan agreed, owners named, no new AI connections to customer records until reviewed | Operating partner and CEO |
| Days 1-30 | Records inventory | System register with history depth, record families and rights status, as metadata only | COO or IT lead |
| Days 1-30 | Guardrails | AI use policy adopted, vendor AI settings recorded, admin accounts under company control | CTO or IT lead |
| Days 31-60 | Use cases | Two use cases chosen, each tied to named record families and an owner | CEO with department heads |
| Days 31-60 | Records check | Completeness and linkage reviewed from field lists and record counts, not exports | COO |
| Days 61-100 | Decisions | Board memo recommending fund, pilot, park or assess for licensing | CEO and CFO |
| Day 100 | Review | Results reported and next-quarter plan set | Operating partner |
Days 1-30: inventory and guardrails#
Days 1-30 build the records inventory and close the easy leaks. The inventory is collected as metadata only: no exports, files or samples leave any system, and nothing goes to vendors or advisors at this stage.
Collect the inventory through short conversations with the people who run each system, not a long form sent to every manager. The help desk lead knows how far back ticket history goes and whether it survived the last migration; the controller knows which ERP modules hold order notes. Mark anything nobody can answer as unknown rather than guessing, and resolve unknowns in month two.
- Systems: every tool that holds business records, from the CRM and help desk to the ERP, field service platform and shared drives.
- History: how many years of records each system still holds and can export.
- Record families: tickets, quotes, jobs, orders, quality records, code reviews, project files, call recordings.
- Linkage: whether records connect a request to the decision and the outcome.
- Rights status: known contract, notice or vendor restrictions, marked clear, restricted or unknown.
- AI exposure: AI tools in use and vendor AI features switched on, with their training settings.
Guardrails that cost little and prevent most surprises#
Guardrails run alongside the inventory and are mostly decisions rather than projects. Adopt a short AI use policy that names approved tools, record each vendor's AI feature and training settings, and move admin accounts for core systems to company-controlled addresses.
The aim is to stop new exposure, not to investigate the past in depth. If the inventory finds a team using an unreviewed AI tool on customer records, pause it, write down what happened and let counsel decide whether anything more is needed. A heavy-handed response in the first month teaches staff to hide tools rather than report them.
Days 31-60: two use cases and the records behind them#
Days 31-60 choose two use cases and check that the records behind them can support them. Two is deliberate: enough to compare, few enough to finish. Each use case should name the record families it depends on and the person who would own it.
Strong candidates sit where the inventory shows deep, linked history. A software company might choose support ticket triage built on tickets linked to engineering issues. A distributor might choose order exception handling built on emails, EDI messages and ERP notes. A trades business might choose estimate review built on estimates, jobs and callbacks.
Check completeness before committing. If the records behind a use case were lost in a migration or live in personal inboxes, the use case is not ready, however attractive it looks on a slide.
Days 61-100: decisions the board can fund#
Days 61-100 turn the work into decisions. The CEO and CFO write a short board memo that gives each option a clear recommendation and the evidence behind it, drawn from the inventory and the records check.
Use the same memo format at every new portfolio company. When the operating partner can compare recommendations side by side, patterns appear across the portfolio, such as several companies parking the same use case for the same records gap, and the sponsor can fix the gap once rather than company by company.
| Decision | Choose it when | Next step |
|---|---|---|
| Fund | The use case has a named owner, complete records and a clear measure of success | Budget and a delivery plan for the next quarter |
| Pilot | Records look adequate but quality or adoption is uncertain | A scoped pilot under reviewed vendor terms |
| Park | Records are incomplete or rights are unclear | Fix the gap and revisit at the next review |
| Assess for licensing | Deep linked history exists that the company does not plan to use itself | A metadata-only fit check approved by the company |
Illustrative: a pest control platform's first 100 days#
Illustrative: a fictional sponsor acquires a regional pest control and lawn care company that runs on a field service platform, a hosted phone system and a shared drive of route sheets. The operating partner and the CEO adopt the template at kickoff.
The inventory shows years of service tickets linked to treatments, follow-up visits and cancellations, plus recorded customer calls whose disclosure language changed twice. The guardrails work finds a marketing contractor running exported customer lists through an AI tool, which the CEO stops.
The two use cases are call summaries for the booking team and a churn warning built on service history. The board funds the call summaries, parks churn until cancellation reasons are cleaned up, and approves a metadata-only licensing assessment of older service history the company has no plan to use itself.
How SourceX fits a 100-day plan#
SourceX fits the last decision in the plan, and only if the board chooses it. The inventory from days 1-30 is the same metadata the Supply step of the SourceX five-step transaction starts from, so a fit check needs no new work and no files.
If the company proceeds, the Rights, Preparation, Approval and Delivery steps follow, and the company approves each one. The SourceX Enterprise Data Value Framework gives the board a plain explanation of why some record families are more likely than others to interest AI developers.
Frequently asked questions
Should the AI plan run separately from the main 100-day plan?
Run it alongside, as one workstream. The records inventory overlaps with IT and finance integration work already under way, so combining them saves effort. Keep the AI plan short and owned by named people so it does not compete with core integration priorities such as reporting and cash.
Who owns the 100-day AI plan?
The portfolio CEO owns delivery, with the COO or IT lead running the inventory and the CFO co-writing the board memo. The operating partner provides the template, reviews progress and connects the company to sponsor resources. Owners should be named at the kickoff, not assigned later.
Do we need outside consultants to run the plan?
Not usually. The company's own leaders can run the inventory and guardrails from a template. Outside help can be useful for specific questions, such as a contract review or a technical look at one use case, but the company should own the plan and its decisions.
What if the inventory shows the company's records are thin?
Then the plan has done its job. Thin or fragmented records point to fixing data capture, admin access and retention before funding AI projects. Record the gaps, park the use cases that depend on them, and set a date to revisit once the basics improve.
Should licensing be on the plan if the board is skeptical?
Keep it as an option, not a commitment. A metadata-only fit check costs the company little and shares no files, and it gives a skeptical board facts instead of assumptions. If the records do not suit licensing, the board drops the option with evidence; if they do, it can decide on its own timetable.
Sources
- In RSM's 2026 middle-market survey, data quality and availability issues were the top inhibitor to AI deployment (34%), followed by security and privacy concerns (30%), legacy systems integration (28%) and talent and skills gaps (28%). Source
Related resources
See if your company qualifies
A short company assessment. No data uploads are needed.