At a glance
- One workflow in supervised production, a measured baseline, and a ranked backlog for day 101 — not a portfolio-wide rollout.
- The week-by-week sequence below is a labeled illustrative example: five workstreams across 100 days, each with an owner and a gate.
- Data readiness at close is uneven, which is why the first three weeks are a diagnostic, not a build.
- Name a workstream owner before choosing a workflow — an unowned workflow stalls at the review gate regardless of how well it's built.
- What gets deferred past day 100 matters as much as what ships: judgment-heavy work and unsettled data stay out of scope.
What should an AI program deliver in the first 100 days after close?
A 100-day AI program should deliver three things: one workflow in supervised production against a measured baseline, a written decision on what to build next, and a named owner who keeps the system running once the deal team leaves. It should not deliver a portfolio-wide rollout or a program that outlives its own business case.
Accordion's survey of 150 private equity operating partners, conducted in May 2026, found that 75% of portfolio companies arrive unprepared or only partially ready for AI deployment at close[1]. We recommend treating that gap as the reason the first three weeks below are a diagnostic against inherited systems, not a build sprint.
Treat the AI workstream like the rest of the value creation plan: one line with an owner, a metric, and a review date, not a side initiative on its own schedule. A result that clears its threshold joins the same EBITDA bridge the fund already tracks; one that doesn't gets one bounded extension at the review gate, then stops at day 100, not carried into the rest of the hold period.
- One workflow in supervised production, reviewed by a person at an agreed rate.
- A before-and-after measurement against the baseline, not an estimate.
- A named owner and a ranked backlog, so day 101 has a start.
The week-by-week 100-day sequence
The sequence below is a labeled illustrative example: a generic shape for the first 100 days, built around five workstreams — diagnose, baseline, build, review, handover. Actual ranges, owners, and gates are set in the written engagement scope, not copied from this table.
| Day range | Workstream | Owner | Exit criterion | Go/no-go gate |
|---|---|---|---|---|
| Days 1–7 (week 1) | Diagnose: access and inventory | Operating partner and CFO | Every target system has a named access owner; read access requested; candidate workflows shortlisted; the workstream owner named. | Go: access filed, workflows shortlisted, owner named. No-go: escalate blocked access. |
| Days 8–21 (weeks 2–3) | Diagnose: workflow selection | Operating partner and process owner | One workflow is scoped to a written boundary: what's in, out, and where exceptions route; the business case and target metric are agreed. | Go: workflow chosen, boundary set, business case agreed. No-go: keep scoping before building. |
| Days 22–35 (weeks 4–5) | Baseline | Process owner | The manual process's current performance is measured on the metric the workflow will be judged on; success and failure thresholds are agreed. | Go: a measured baseline and agreed thresholds exist. No-go: don't build without them. |
| Days 36–56 (weeks 6–8) | Build: design and shadow run | Implementation lead | The system shadows the manual process without acting, so output can be compared case by case. | Go: error and disagreement rates meet the agreed threshold. No-go: redesign or extend the shadow run. |
| Days 57–77 (weeks 9–11) | Build: supervised production | Process owner and implementation lead | The system acts on live work; a person reviews every case at an agreed rate, with overrides logged. | Go: error and override rates hold at threshold for two weeks. No-go: pause, return to the shadow run. |
| Days 78–91 (weeks 12–13) | Review | Operating partner and CFO | Results are measured against the baseline and reported against the business case agreed in days 8–21. | Go: criteria met, result reported up. No-go: use the remaining days as one bounded extension. |
| Days 92–100 (weeks 14–15) | Handover | Named portfolio-company owner | Documentation, runbooks, and the evaluation set are handed over; the next candidate is scoped for day 101. A workflow still short of its threshold stops here instead. | Go: owner signs off, next workflow scoped. No-go: stop; document why and scope the next-ranked candidate. |
What should be deferred until after day 100?
Defer anything that needs more baseline history than you have, any workflow touching a customer or vendor before a person signs off, and judgment calls beyond a documented rule. Accordion's May 2026 survey of 150 PE operating partners found only 14% had moved past back-office automation into genuine decision intelligence[1] — a reasonable split for 100 days.
- Rolling the first workflow out to other portfolio companies — prove it in one place first.
- Any workflow whose data can't yet support the days 22–35 baseline measurement.
- A second workflow running in parallel before the first clears its threshold.
Which AI use cases can reach production inside 100 days?
The use cases that reach production in 100 days share three traits: real volume flowing through one system, a metric someone already tracks by hand, and exceptions a person can review at a defined rate. Good first candidates include KPI reporting, multi-ERP close tracking, and collections sequencing, covered on our private equity industry page.
Accordion's May 2026 survey of 150 PE operating partners found 83% report at least one agentic AI use case deployed or in pilot in a portfolio-company finance function[1]. What's rare is the harder kind of use case — judgment beyond a known rule — which is why the sequence above scopes one workflow first.
- High-volume, low-judgment workflows: invoice-to-PO matching, KPI consolidation across entities, dunning sequencing by invoice age.
- Workflows with an existing, measurable baseline and an exception path a person already handles today.
What data and system access is needed in week one?
Week one needs read access to whatever system holds the candidate workflow's data — whatever ERP came with the company at close — plus a named approver on the portfolio-company side. Write access waits until the design stage; week one is about seeing the data, not acting on it.
Microsoft's documentation for Dynamics 365 Finance and Operations combines individual security permissions into privileges and duties, and each privilege specifies a level of access, such as Read or Delete, for the entry point it covers[2]. Request the narrowest role that lets the diagnostic see the data.
| System or data | Access level needed | Who approves it | Why it gates week one |
|---|---|---|---|
| The candidate workflow's system of record (ERP, billing platform, or BI tool) | Read-only | Portfolio-company IT or controller | Without it, the diagnostic can't see real volume or the exception mix. |
| Historical transactions for the baseline period, and metric definitions across entities | Read-only extract | Process owner or controller | The baseline and any KPI workflow need real history, not an estimate. |
| A sandbox or test environment, where one exists | Full, sandbox only | Portfolio-company IT | Lets the build stage test without touching live records. |
Who at the portfolio company should own the AI workstream?
One named person at the portfolio company, not a committee. We recommend the CFO or controller for finance workflows, reporting to the sponsoring operating partner. That person is named as part of the days 1–7 access workstream below, and makes the go/no-go call at each gate that follows — the same accountability any value-creation initiative gets.
FTI's 2026 PE AI Radar (200 fund/operating leaders, published May 2026) found talent remains the primary constraint to scaling AI adoption, cited by 35% of respondents[3]. Name the owner before choosing the workflow.
How do you report 100-day AI progress to the investment committee?
An investment-committee report has four entries: the workflow and business case by day 21; the baseline and thresholds by day 35; the shadow run's results by day 56; and the production result, go/no-go decision, and next candidate by day 91.
FTI Consulting's 2026 Private Equity AI Radar (published May 2026, 200 fund and operating leaders) found 95% of funds report AI initiatives meeting or exceeding their business case criteria[3].
Where to start: a first scoped workflow
The 360° AI Blueprint is a diagnostic roadmap across your entire organization. It pinpoints where performance leaks, where AI creates the biggest impact, and what each initiative returns — quantified and sequenced from quick wins to long-term moves. It's where we start when the starting point isn't clear; if you already have a specific project in mind, we jump straight in.
A 360° AI Blueprint run against the systems inherited at close produces the ranked backlog the day-100 gate needs. A fund that already ranked candidates at the platform level can start straight from AI agent implementation for the workflow chosen.
Before committing to one workflow, price the status quo the way you'd price any capital decision. Our guide to how to calculate AI ROI before you build walks through that method; what a 360° AI Blueprint contains explains how the diagnostic ranks candidates.
- Which inherited system holds the workflow's data, and who can grant read access this week.
- Whether a baseline already exists, or needs about two weeks to measure.
- Who signs the go/no-go call at each gate — named once, not decided later.
When a 100-day plan isn't the right frame
A fixed 100-day clock is the wrong frame in a few common situations:
- If a workflow's answer is just a lookup or a fixed calculation an existing system can already do, there's no diagnostic-to-build sequence to run: turn on the module you already own instead of building an agent for it.
- The business is still being stabilized post-close; an AI gate on an unstable org chart adds a failure mode nobody needed.
- There's no realistic way to measure a baseline in two weeks, or the fund's review cadence doesn't allow a day-100 decision; align the gate to the calendar that exists.
Frequently asked questions
What should an AI program deliver in the first 100 days after close?
One workflow in supervised production against a measured baseline, a written decision on what to build next, and a named owner who keeps it running once the deal team moves on — not a portfolio-wide rollout, and not a program that outlives the case that funded it.
Which AI use cases can reach production inside 100 days?
Workflows with real volume flowing through one system, a metric someone already tracks by hand, and exceptions a person can review at a defined rate. KPI consolidation, multi-ERP close tracking, and collections sequencing are good first candidates; judgment-heavy decision work waits until after day 100.
What data and system access is needed in week one?
Read-only access to the system holding the candidate workflow's data — whatever ERP came with the company, whether NetSuite, Sage Intacct, or another system — plus historical transactions for the baseline period and a named approver. Write access waits for the design stage, once the boundary is scoped.
Who at the portfolio company should own the AI workstream?
One named person, we recommend the CFO or controller for finance workflows, reporting to the sponsoring operating partner. That person approves access, reviews the shadow run, and makes the go/no-go call. Name that owner before choosing the workflow: an unowned workflow stalls at the review gate.
What should be deferred until after day 100?
Rolling the first workflow out to other portfolio companies, any workflow whose numbers are still unreliable this soon after close, decisions that call for judgment beyond a documented rule, and starting a second workflow before the first clears its threshold. Prove the high-volume case first.
How do you report 100-day AI progress to the investment committee?
A four-part packet tied to the sequence's gates: the workflow and business case by day 21, the baseline and thresholds by day 35, the shadow run's results by day 56, and the production result, the go/no-go decision, and the next candidate by day 91. Route investor-facing material through compliance first.
Related reading
- The 360° AI Blueprint deliverable is "the initiatives worth building: AI and automation candidates ranked by ROI, cost, and time-to-value."
- What is a 360° AI Blueprint? covers what the diagnostic contains and how the ranking works.
- Our framework for how to calculate AI ROI before you build prices the status quo before you commit to a workflow.
- How to scope an AI automation pilot is the worksheet behind the baseline metric and the success and failure thresholds the build stage above sets.
Sources
- Accordion, "The PE AI Adoption Benchmark" (2026) (accessed 2026-09-30)
- Microsoft, "Security architecture - Finance & Operations" (accessed 2026-09-30)
- FTI Consulting, "2026 Private Equity AI Radar" (2026) (accessed 2026-09-30)
More in AI strategy
AI Agents vs. RPA for Hotel Back-Office Operations: When Bots Break and What Replaces Them
AI agents vs RPA for hotel back-office operations: a decision table for OTA statement matching, folio audit and multi-property AP, plus how to phase bots.
Hotel Management CompaniesWhat a 360 AI Blueprint Costs for a Food & Beverage Manufacturing Back Office
What drives food and beverage manufacturing AI cost in the back office: a table of plants, co-packers, deductions and traceability scope, plus how to budget.
Food and Beverage ManufacturersAI Implementation Cost for Multi-Rooftop Dealer Groups: What Changes When You Add a Store
AI implementation cost for multi-rooftop dealer groups: a driver table showing how DMS count, OEM portals, build type and upkeep move cost per added store.
Auto Dealerships