Skip to main content
How Can We Help?Contact

Private Equity & Portfolio Companies

Test the data before you sign, and prove the AI value before you sell.

We work with deal teams, operating partners and portfolio company leaders from diligence to exit. Revenue, customer and margin data are tested against the model before signing; the first 100 days put one set of KPIs on the company’s own systems; AI initiatives are tied to lines in the P&L and measured in production; and the exit data room is ready before the buyer asks.

Talk through your deal or portfolio See three results by value lever

Buyout and growth funds · Operating and value creation teams · Portfolio company CEOs and CFOs · Add-ons and carve-outs

A deal timeline with four milestones; above them a data room, a 100-day plan, an EBITDA bridge and the buyer’s checklist light in turn as a pulse moves from signing to exit1001234
  1. 1DiligenceRevenue by customer tied to the ledger before signing
  2. 2First 100 daysOne KPI pack, defined once, reported monthly
  3. 3Value creationPricing uplift in run-rate EBITDA, measured two quarters
  4. 4ExitThe buyer’s data questions answered from one room
Illustrative: one deal, from the seller’s data room to the buyer’s.

The red flags between close and exit

Five findings we see in deal after deal, set out the way a diligence report sets them out: what the deal team sees, what the company sees, the cause, and the value at risk.

  1. 1.1

    Red flag

    The numbers in the CIM cannot be rebuilt from the systems

    The deal team sees
    Revenue by customer, churn and margin by product arrive as spreadsheets that do not tie to the ledger.
    The company sees
    The first board pack takes weeks, and its numbers differ from the deal model.
    The cause
    Metrics assembled by hand from several systems, with definitions that changed over the years.
    Value at risk
    A price paid on numbers nobody can reproduce, and a value creation plan built on the wrong baseline.
    See Folder 03: the KPI dictionary

  2. 1.2

    Red flag

    Every portfolio company reports the same KPI a different way

    The deal team sees
    Twelve monthly packs in twelve formats, so the portfolio cannot be compared or rolled up.
    The company sees
    A finance team spending a week each month rebuilding numbers for the board and the fund.
    The cause
    No shared definitions, and no feed from the companies’ systems to the fund.
    Value at risk
    Problems seen a quarter late, and LP reporting that takes weeks to assemble.
    See Folder 04: the work across the portfolio

  3. 1.3

    Red flag

    AI is in the plan, not in the P&L

    The deal team sees
    AI initiatives in the value creation plan with EBITDA attached, and no owner, data or date.
    The company sees
    Pilots that never reach production while their run-rate is still counted.
    The cause
    Initiatives chosen from slides rather than from the data the company actually holds.
    Value at risk
    A multiple paid at entry, and asked for at exit, on EBITDA a buyer will not credit.
    See Folder 03: the bridge priced three ways

  4. 1.4

    Amber flag

    The exit data room is rebuilt from scratch

    The deal team sees
    The buyer’s data requests answered by hand, months before signing.
    The company sees
    Management pulled off the business to assemble data for the next owner.
    The cause
    Nothing kept through the hold in a form a buyer’s diligence can use.
    Value at risk
    Delay, price chips and larger escrows when the numbers cannot be shown.
    See Folder 04: the work before exit

  5. 1.5

    Amber flag

    Privacy and AI risks surface in confirmatory diligence

    The deal team sees
    Customer data, consents and AI tools found late, with no inventory to check them against.
    The company sees
    Remediation landing in the first 100 days instead of the value creation plan.
    The cause
    No data map, no AI inventory, and no owner for either.
    Value at risk
    Special indemnities, escrows and insurance exclusions for issues found before signing.
    See Folder 05: the schedules

Three results, filed by value lever

No portfolio-company engagement is published yet. These three results are from operating companies, filed by the lever each would sit under in a value creation plan, at full length and with the client’s own figures; the clients are described, not named.

Exhibit 2.1

Revenue lever

A national quick-service chain

Loyalty offers chosen by who they will move, at less discount

Offers were broad-based rather than tuned to a member’s lifecycle or to how they visited, so national offers did not always produce the best local outcome. First-party transactions and loyalty data were combined with public, aggregated spend signals to find where a targeted incentive would convert best, with eligibility kept consistent across the app, mobile order-ahead and the store, and clean A/B tests with guardrails so that promotional spend went down while loyalty growth and repeat visits went up.

Scope of work

  • Lifecycle segments and propensities: new, routine and lapsing members, scored for conversion, frequency, basket lift and offer fatigue.
  • Offer selection with caps: the incentive, its depth and its timing, with margin and frequency caps enforced.
  • Spend reallocation: promotional spend moved toward the cohorts and offers that performed best.

Key figures

  • +14%loyalty sign-ups where offers were targeted, by daypart and market
  • +11%repeat visits from active members on tuned offers
  • −8%discount spent for each incremental visit

Reported by the client.

Case study: The QSR loyalty rewards case

Exhibit 2.2

Cost-to-serve lever

A global manufacturer of electronics

A virtual agent that resolves support issues before an agent is needed

A large share of inbound support issues were repeat questions the systems could answer: status, entitlement and basic troubleshooting. The assistant had to work across several service systems through their APIs, keep the context, offer a few high-accuracy options rather than walls of text, and hand off cleanly to the service desk when self-service was not enough.

Scope of work

  • An omni-channel virtual agent: chat and voice, so support starts wherever the user already is.
  • Top-three guided responses: the three most likely fixes from the question and the user’s history, narrowed interactively.
  • Contextual hand-off: the full interaction passed to the service desk, so the agent never re-asks.
  • Live system data: device and service information read from the manufacturer’s own servers.

Key figures

  • −37%Tier 1 live-agent contacts, for the issue types covered
  • +28%resolution on first contact
  • 22%faster time to answer for customers who started with the agent

Reported by the client.

Case study: The technical-support virtual agent case

Exhibit 2.3

Operations lever

A food manufacturer and distributor supplying major grocers

A planning and routing layer on top of the existing planning system

The business wanted to see who it was really supplying, from where and on which routes, across supermarkets, distribution centers and warehouse deliveries. Its planning system handled demand and production well but did not use store-level pull, route performance or lead-time variability, and commercial reality had to be connected to plant output, inventory, truck capacity and driver windows without disrupting it.

Scope of work

  • Supply-relationship mapping: who buys what, from where and how often, reconciled.
  • Cleaner planning signals: demand, lead-time and location data fed into the existing planner.
  • Route optimization: deliveries sequenced and loads consolidated within delivery windows.
  • Planner dashboards: planned against actual by customer and location.

Key figures

  • −12%delivery miles per fulfilled order
  • −18%stockout incidents at key supermarket partners
  • +9%plan adherence; the planning system was never replaced

Reported by the client.

Case study: The supply mapping and routing case

Price the value creation plan the way a buyer will

Start with one company’s EBITDA bridge, read three ways. Then open the KPI dictionary: seven numbers every portfolio company reports, each defined once, with what to settle before the next board pack.

One EBITDA bridge, priced three ways

A portfolio company with 40.0 million of EBITDA and five data and AI initiatives in its value creation plan. Choose how the bridge is read: the exit EBITDA, and what it is worth at 11 times, change with it.

Read the bridge

Run-rate EBITDA added+5.3 million

Worth at 11 times58.3 million

What the plan, and often the management presentation, claims.

Run-rate EBITDA added+3.1 million

Worth at 11 times34.1 million

What is actually running today.

Run-rate EBITDA added+2.6 million

Worth at 11 times28.6 million

What a buyer’s quality-of-earnings review is likely to credit: 29.7 million of the plan’s value still rests on initiatives without two quarters of numbers.

Illustrative figures, checked line by line. The multiple is an assumption; the point holds at any multiple.

EBITDA bridge, millions of dollars

The five initiatives in the bridge
InitiativeRun-rate EBITDA, millionsStatus in the plan
Pricing analytics+1.6Measured for three quarters
Procurement spend+1.0Measured for two quarters
Inventory and markdowns+0.5Live for one quarter
Service assistant+0.9In pilot
Sales next-best-action+1.3In the plan only

The KPI dictionary

3.1KPI dictionary · one definition for every company

Adjusted EBITDA

Calculated as

  1. Reported. EBITDA as the accounts show it.
  2. With add-backs. Plus one-off costs such as severance and deal fees, and the full-year effect of an add-on bought during the year.
  3. With run-rate adjustments. Plus savings and initiatives counted as if realized for the full year.

Read by

The lenders’ covenants, the board, and every buyer at exit.

Drawn from

the general ledger, the add-back schedule, the initiative tracker

Worked through

Reported EBITDA is 36.0 million. One-off costs of 2.0 million and a full year of an add-on bought in June, another 2.0 million, make 40.0. Run-rate savings not yet realized add 1.5: adjusted EBITDA is 38.0, 40.0 or 41.5 million, depending on what is added back.

Settle in the reporting manual

Which add-backs, and does anything count before it is realized?

3.2KPI dictionary · one definition for every company

Net revenue retention

Calculated as

  1. All movement. Revenue this year from last year’s customers, after expansion, contraction and churn.
  2. Volume only. With price increases taken out, so growth from selling more is seen apart from charging more.
  3. By what. By customer, by contract, or by product line.

Read by

The deal model, the growth story at exit, and where the sales team spends its time.

Drawn from

billing, the CRM, the general ledger

Worked through

Last year’s customers paid 50.0 million. This year they pay 6.0 million more in expansion and 4.0 million less from churn and contraction: 52.0 million, a retention of 104%. If 2.5 million of the expansion is price alone, retention by volume is 99%.

Settle in the reporting manual

All movement or volume only, and counted by customer or by contract?

3.3KPI dictionary · one definition for every company

Gross margin

Calculated as

  1. Product cost only. Revenue less the direct cost of what was sold.
  2. With delivery. Freight and warehousing counted as cost of sales.
  3. With service. Implementation and support staff counted too.

Read by

Pricing, the mix the sales team is paid on, and comparisons across the portfolio.

Drawn from

the general ledger, the cost and inventory system, timekeeping

Worked through

Revenue of 100.0 million and product cost of 55.0 million give 45%. With 4.0 million of freight it is 41%, and with 3.0 million of implementation staff it is 38%.

Settle in the reporting manual

Which costs sit above the gross margin line, the same in every company?

3.4KPI dictionary · one definition for every company

Cash conversion

Calculated as

  1. Before capital spending. Operating cash flow as a share of EBITDA.
  2. After capital spending. Operating cash flow less capital spending, as a share of EBITDA.
  3. Which working capital. With or without the movement in deferred revenue and supplier payments timed around quarter end.

Read by

The debt the company can carry, dividends to the fund, and the cash story at exit.

Drawn from

the cash flow statement, the fixed asset register, the general ledger

Worked through

EBITDA of 40.0 million, working capital up 4.0 million and capital spending of 6.0 million: conversion is 90% before capital spending (36.0 over 40.0) and 75% after it (30.0 over 40.0).

Settle in the reporting manual

Before or after capital spending, and which working capital movements count?

3.5KPI dictionary · one definition for every company

Customer acquisition payback

Calculated as

  1. Which spend. All sales and marketing, or only the share spent winning new customers.
  2. Which margin. Months to recover the cost from gross margin, not from revenue.
  3. New or expanded. New customers only, or expansions counted too.

Read by

How much to spend on growth, and whether growth is buying itself.

Drawn from

the CRM, marketing spend by channel, billing

Worked through

Sales and marketing cost 6.0 million and win 300 new customers: 20,000 dollars each. Each brings 1,000 dollars of gross margin a month, so payback is 20 months. Count only the 60% of spend that goes on new customers and it is 12 months.

Settle in the reporting manual

Which spend, which margin, and which customers count as new?

3.6KPI dictionary · one definition for every company

Revenue per employee

Calculated as

  1. Employees only. Revenue divided by the people on the payroll.
  2. Full-time equivalents. Part-time staff counted as fractions, and contractors counted in.
  3. Which revenue. Total revenue, or revenue less pass-through costs.

Read by

Productivity targets, and the case for automation in the value creation plan.

Drawn from

payroll and the HR system, contractor invoices, the general ledger

Worked through

Revenue of 120.0 million with 600 employees is 200,000 dollars each. Count 150 full-time contractors as well and it is 160,000.

Settle in the reporting manual

Payroll only or full-time equivalents, and which revenue?

3.7KPI dictionary · one definition for every company

Days sales outstanding

Calculated as

  1. Period end. Receivables at quarter end, against the quarter’s revenue.
  2. Average. Average receivables over the quarter.
  3. Which receivables. Billed only, or unbilled work counted too.

Read by

Working capital targets, the lenders, and the collections team.

Drawn from

the receivables ledger, billing, the general ledger

Worked through

Quarterly revenue is 27.0 million. Receivables of 18.0 million at quarter end give 61 days (18.0 over 27.0, times 91). Average receivables of 15.0 million give 51 days.

Settle in the reporting manual

Period end or average, and are unbilled receivables in?

Which number would a buyer struggle to rebuild today? Bring it to a 30-minute call, and we will show you how to define it once and feed it from the systems for every company in the portfolio.

Test one number before the next board pack

The work, from signing to exit

Four stages, the way a deal runs rather than the way the systems are built. Each document says how it shows up in the numbers; where a published result exists, it is linked.

Stage 1Before signing

Between the LOI and signing

  • 4.1

    Data diligence alongside the financial and commercial work

    Revenue, customers, churn and margin rebuilt from source systems and tied to the ledger, with each gap between the CIM and the systems listed and what it is worth.

    Shows up as: Findings priced before signing

    First step: AI & Data Assessment Sprint

  • 4.2

    An AI and data read on the investment thesis

    Where data and AI could move EBITDA in this business, sized against the data the company actually holds, so the value creation plan starts from evidence rather than slides.

    Shows up as: Initiatives in the plan with the data to support them

    First step: Executive AI Workshop

Stage 2The first 100 days

From close to the first full board

  • 4.3

    One KPI pack from the systems, not from spreadsheets

    The board’s and the fund’s KPIs defined once and fed from the company’s own systems, with lineage to the ledger, live by the first full board meeting.

    Shows up as: Days to the monthly board pack

    First step: Data Platform Accelerator

  • 4.4

    A data foundation sized to the plan

    Only the data the value creation plan needs, modeled first; for a buy-and-build, the way each add-on joins is designed in from the start.

    Shows up as: Plan initiatives with the data they need

    First step: Data Platform Accelerator

Stage 3The hold

Value creation, measured

  • 4.5

    AI initiatives taken to production and measured

    Each initiative built against a line already in the P&L, with a baseline and a control group, so the run-rate is measured rather than estimated.

    Shows up as: Run-rate EBITDA measured for two quarters

    Case study: The technical-support virtual agent case

  • 4.6

    Pricing, procurement and operations analytics

    Price leakage, off-contract spend and inventory by item and location, where most portfolio companies find their first measurable EBITDA, layered on the systems they already run.

    Shows up as: Price realization, spend under contract and stockouts

    Case study: The supply mapping and routing case

Stage 4Before exit

From the decision to sell to signing

  • 4.7

    An exit data room that answers the buyer’s questions

    The KPIs, cohorts and AI evidence a buyer’s diligence asks for, kept current from the systems through the hold, so the sell-side data room is assembled in days.

    Shows up as: Days to answer the buyer’s data requests

    First step: Data Platform Accelerator

  • 4.8

    AI and privacy evidence for the disclosure schedules

    The AI inventory, data map, consents, risk assessments and incident log a buyer and its insurer will ask for, with an owner for each.

    Shows up as: Open diligence findings at signing

    First step: AI Governance Quickstart

Across the portfolio

  • A portfolio data feed to the fund

    Each company’s KPIs on one definition, delivered monthly to the operating team, and ready for the LP reporting formats investors now ask for.

    First step: Data Platform Accelerator. Backed by Data Engineering & Platforms.

  • One AI policy for every company

    A light, shared AI policy, inventory and approval path adopted company by company, so the fund can answer an LP or a buyer about AI anywhere in the portfolio.

    First step: AI Governance Quickstart. Backed by Strategy & Governance.

  • Management teams who can run it

    Training by role for each company’s leaders and analysts, so the tools are still in use when the next owner arrives.

    First step: Adoption Review. Backed by AI & Data Literacy.

The deal or the company you have in mind sits somewhere else on this timeline? Bring it to a 30-minute call.

Talk through one deal or company

The schedules a buyer, an examiner or an LP will read

The rules that reach the data and AI in a deal, grouped by what they bind: the transaction, the portfolio company, and the fund. Each schedule gives the question you will be asked and the evidence to keep in the data room, and links to its source.

Rules that reach data and AI in private equity: what each binds, its status, the question it raises and the evidence to keep
Schedule and what it bindsStatus at 14 October 2026The question you will be askedEvidence to keep in the data room
The transaction
Schedule 5.1DOJ Data Security Program (28 CFR Part 202) (opens in a new tab)US companies, funds included, that give a country of concern or a covered person access to bulk US sensitive personal data, including through investors, vendors or staff.In force since Apr 8, 2025. Compliance programs, audits and reporting for restricted transactions since Oct 6, 2025.From a buyer: has the company given any country of concern or covered person access to bulk sensitive personal data?What sensitive data the company holds and in what volume (health and financial data count from 10,000 people), who can reach it, and any investors, vendors or staff linked to China, Russia, Iran, North Korea, Cuba or Venezuela.
The portfolio company
Schedule 5.2State privacy laws, and California’s regulations on risk assessments (opens in a new tab)Portfolio companies covered by the privacy laws of the 20 states that have one, and by California’s regulations.Indiana, Kentucky and Rhode Island joined on Jan 1, 2026. California: notices for automated decisions from Jan 1, 2027; risk assessment details for 2026 and 2027 due Apr 1, 2028.From a buyer: does the company comply with privacy laws and its own notices, and has it had an incident it did not report?A data map, consent and opt-out records, risk assessments, vendor contracts and the incident log.
Schedule 5.3Automated decisions about people: New York City, Illinois, California and Colorado (opens in a new tab)Portfolio companies that use automated tools to hire, promote or decide about people.New York City’s bias audits since Jul 5, 2023; Illinois notices since Jan 1, 2026; California’s and Colorado’s rules from Jan 1, 2027.From a buyer: which AI tools make or shape decisions about people, and does each have the audits and notices the law requires?An AI inventory with an owner for each tool, bias audit results, and the notices given.
The fund
Schedule 5.4SEC Regulation S-P, as amended in 2024 (opens in a new tab)Registered investment advisers, private fund advisers included.Advisers with 1.5 billion dollars or more under management since Dec 3, 2025; all others since Jun 3, 2026.From an examiner or an LP: can you detect, contain and report unauthorized access to investor information, and tell the people affected within 30 days?An incident response program, oversight of service providers (who must tell you within 72 hours), and a map of where investor information sits.
Schedule 5.5The SEC’s examination focus on AI claims (opens in a new tab)Registered advisers that describe AI in marketing, Form ADV or communications with investors.Fiscal 2026 priorities, published Nov 17, 2025: examiners compare what advisers say about AI with what they use.From an examiner: does the AI you describe to investors, value creation included, exist and do what you say it does?A record of each AI tool and initiative described to investors, what it does, and the results behind any figure quoted.
Schedule 5.6Form PF amendments (opens in a new tab)SEC-registered advisers with 150 million dollars or more in private fund assets.Compliance date moved to Jul 1, 2027 on Aug 31, 2026, while the SEC and CFTC consider an April 2026 proposal to change or remove parts of the amendments.From the SEC: more detailed fund, portfolio and financing data.Fund and portfolio data held in one place with its definitions, so the filing is produced rather than assembled.
Schedule 5.7ILPA Reporting and Performance Templates (opens in a new tab)Not a law: voluntary templates that many LPs now ask for.Reporting Template from Q1 2026 for funds in their investment period and funds started from Jan 1, 2026; first Performance Templates from Q1 2027.From an LP: fees from gross to net, itemized expenses and carried interest, and IRR, TVPI and MOIC with and without the subscription line.Cash flows tagged by transaction type, and fund expenses tied to the general ledger.

For orientation only, not legal advice: other federal, state and foreign rules may apply, and what applies depends on the fund, the deal, the company and its data. Checked October 14, 2026.

Screen an AI initiative the way an investment committee would

Select an initiative to see the thesis, where it lands in EBITDA, the screening questions, the rules to clear, the conditions to scale, and who would usually approve it.

6.1Revenue

Pricing analytics and price realization

The thesis
Finds where list prices, discounts and mix leak margin, by customer and product, and recommends price moves.
Where it lands in EBITDA
Gross margin, through price realization.

Screening questions

  • Uses customer personal dataIn part
  • Decides about employees or applicantsNo
  • Touches fund or investor dataNo
  • Faces customers directlyNo
  • Needs data from several companiesNo

Rules to clear

  • Competition law, if the tool draws on competitors’ prices, as California’s AB 325 now spells out.
  • State limits on personalized prices, if prices are set per customer from personal data.

Conditions to scale

  • A commercial lead approves every price change.
  • Results measured against a control group of customers or products.
  • Run-rate counted only after two full quarters.

Approved by (suggested)

Operating partner

First step: AI & Data Assessment Sprint

6.2Revenue

Next-best-action for the sales team

The thesis
Suggests which customers to contact, and with what offer, from purchase history and signals.
Where it lands in EBITDA
Revenue, through win rate and cross-sell.

Screening questions

  • Uses customer personal dataYes
  • Decides about employees or applicantsNo
  • Touches fund or investor dataNo
  • Faces customers directlyNo
  • Needs data from several companiesNo

Rules to clear

  • State privacy laws on profiling and targeted marketing.
  • Call and text consent rules, if it triggers outreach.

Conditions to scale

  • Sellers decide; the tool suggests.
  • Win rates compared with a control group.
  • Customer opt-outs carried into every suggestion.

Approved by (suggested)

Function lead

First step: AI & Data Assessment Sprint

6.3Cost base

A customer service assistant

The thesis
Answers routine questions, looks up the account, and hands off to an agent with the context.
Where it lands in EBITDA
Cost to serve, through fewer live contacts.

Screening questions

  • Uses customer personal dataYes
  • Decides about employees or applicantsNo
  • Touches fund or investor dataNo
  • Faces customers directlyYes
  • Needs data from several companiesNo

Rules to clear

  • Bot disclosure laws in California, Maine and Utah.
  • State privacy and recording laws for transcripts.

Conditions to scale

  • Answers only from approved content and account data.
  • Says it is an assistant, and offers a person.
  • Savings counted on the issue types it covers.

Approved by (suggested)

Function lead

First step: AI & Agent Pilot. Case study: The technical-support virtual agent case

6.4Cost base

Procurement spend analytics

The thesis
Classifies spend, finds off-contract buying and duplicate suppliers, and compares the prices paid.
Where it lands in EBITDA
Cost of goods and overhead, through spend under contract.

Screening questions

  • Uses customer personal dataNo
  • Decides about employees or applicantsNo
  • Touches fund or investor dataNo
  • Faces customers directlyNo
  • Needs data from several companiesIn part

Rules to clear

  • Competition law, if supplier prices are shared between portfolio companies that compete.

Conditions to scale

  • Savings booked only when the new price shows in invoices.
  • Category owners confirm every reclassification.
  • Supplier data shared across companies only in aggregate.

Approved by (suggested)

Function lead

First step: AI & Data Assessment Sprint

6.5Working capital

Demand forecasting and inventory

The thesis
Forecasts demand by item and location, and sets stock levels and orders.
Where it lands in EBITDA
Working capital and lost sales, through inventory turns; it shows in cash before EBITDA.

Screening questions

  • Uses customer personal dataNo
  • Decides about employees or applicantsNo
  • Touches fund or investor dataNo
  • Faces customers directlyNo
  • Needs data from several companiesNo

Rules to clear

  • A forecast without personal data falls outside most rules.

Conditions to scale

  • Accuracy measured on the company’s own history, weighted by sales.
  • Planner overrides allowed and logged.
  • Stock targets changed in steps, with service levels watched.

Approved by (suggested)

Function lead

First step: AI & Data Assessment Sprint. Case study: The supply mapping and routing case

6.6Cost base

Finance close and board pack automation

The thesis
Drafts reconciliations, variance commentary and the monthly board pack from the ledger.
Where it lands in EBITDA
Finance cost, and days to close.

Screening questions

  • Uses customer personal dataNo
  • Decides about employees or applicantsNo
  • Touches fund or investor dataIn part
  • Faces customers directlyNo
  • Needs data from several companiesNo

Rules to clear

  • No AI-specific rule; the controls the auditors rely on still apply.

Conditions to scale

  • A controller reviews every reconciliation and comment.
  • Figures tie to the ledger before anything is drafted.
  • Every change logged for the auditors.

Approved by (suggested)

Function lead

First step: AI & Agent Pilot

6.7Deal team

Contract review in the data room

The thesis
Reads customer and supplier contracts in diligence for change of control, exclusivity, pricing terms and assignment.
Where it lands in EBITDA
Deal risk found before signing, and terms to fix in the first 100 days.

Screening questions

  • Uses customer personal dataIn part
  • Decides about employees or applicantsNo
  • Touches fund or investor dataNo
  • Faces customers directlyNo
  • Needs data from several companiesNo

Rules to clear

  • The data room’s confidentiality terms and the NDA: where documents may be processed.
  • Privacy laws, for personal data inside the contracts.

Conditions to scale

  • Runs inside the fund’s or the adviser’s environment; no document trains a model.
  • Counsel confirms every flagged clause.
  • Each finding logged against its document and page.

Approved by (suggested)

Operating partner

First step: AI & Agent Pilot. Case study: The contract compare-and-comply case

6.8Hiring and staffing

Applicant screening and workforce planning

The thesis
Ranks applicants, or flags roles and people in a restructuring.
Where it lands in EBITDA
Hiring cost and time to fill, or payroll.

Screening questions

  • Uses customer personal dataNo
  • Decides about employees or applicantsYes
  • Touches fund or investor dataNo
  • Faces customers directlyIn part
  • Needs data from several companiesNo

Rules to clear

  • New York City Local Law 144: a bias audit and notice before use.
  • Illinois notices since Jan 1, 2026; California and Colorado from Jan 1, 2027.
  • Layoff notice laws apply whatever the tool says.

Conditions to scale

  • A person decides; the tool ranks or flags.
  • Selection rates checked by group before use.
  • Applicants and employees told when a tool is used.

Approved by (suggested)

Board or investment committee

First step: AI Governance Quickstart

6.9Fund

Portfolio monitoring and KPI collection

The thesis
Collects each company’s KPIs on one definition and flags companies drifting from plan.
Where it lands in EBITDA
Earlier action on companies behind plan.

Screening questions

  • Uses customer personal dataNo
  • Decides about employees or applicantsNo
  • Touches fund or investor dataYes
  • Faces customers directlyNo
  • Needs data from several companiesYes

Rules to clear

  • Regulation S-P, for any investor information it holds.
  • The information rights in each company’s shareholder agreement.

Conditions to scale

  • Definitions agreed with each company’s CFO.
  • Every figure traceable to the company’s ledger.
  • Access limited by deal team and by company.

Approved by (suggested)

Operating partner

First step: Data Platform Accelerator

6.10Fund

Drafting LP reports and due diligence questionnaires

The thesis
Drafts quarterly letters and questionnaire answers from approved fund data and past answers.
Where it lands in EBITDA
Investor relations time; no direct EBITDA effect.

Screening questions

  • Uses customer personal dataNo
  • Decides about employees or applicantsNo
  • Touches fund or investor dataYes
  • Faces customers directlyIn part
  • Needs data from several companiesIn part

Rules to clear

  • The SEC Marketing Rule, for anything that reaches prospective investors.
  • The SEC’s examination focus on accurate AI claims.
  • Regulation S-P, for investor information.

Conditions to scale

  • Answers drawn only from approved, dated sources.
  • Compliance reviews every performance figure and claim.
  • No investor data leaves the firm’s environment.

Approved by (suggested)

Board or investment committee

First step: AI Governance Quickstart

A first screen to take into the investment or portfolio review, not a legal opinion. What applies depends on the company, its data, where it operates and how the tool is used; counsel and the company’s own owners decide. The schedules in full: Folder 05.

Suggested approvals. More than one may be needed: the operating partner backs the initiative, the company’s board approves anything that decides about people or investors, and counsel decides what has to be disclosed.

  • Board or investment committee: for tools that decide about people, or that reach investors.
  • Operating partner: with the company CEO, for initiatives counted in the value creation plan.
  • Function lead: a named owner, access controls and a quarterly check.

An initiative already in a value creation plan, or one a management team is pushing? Thirty minutes is usually enough to see what it would take to measure it, and what it needs to clear.

Screen one initiative with us

Three ways in, set out as indicative terms

Three options for three different readers, side by side the way a term sheet sets them out. The first two are our assessment and executive offers, sized to one company or one operating team, and neither commits you to the next.

  1. Track A

    Data diligence for one deal

    Who it suits
    A deal team between the LOI and signing, or in the first 30 days after close.
    Timing
    Two weeks, one company
    What you receive
    The company’s revenue, customer and margin data rebuilt from source and tied to the ledger; the gaps against the CIM and the model, each with what it is worth; and the data work the value creation plan will need, in order.
    Scope diligence for a deal

  2. Track B

    AI priorities across the portfolio

    Who it suits
    An operating partner choosing where AI goes first across several companies.
    Timing
    Several hours, one operating team
    What you receive
    The operating team works through which companies and which levers AI should reach first, against the data each company actually holds, and leaves with a ranked shortlist and one first move.
    Plan a session for the operating team

  3. Track C

    Data Maturity Assessment

    Who it suits
    A portfolio company CFO or CIO who wants a quick read first.
    Timing
    Six minutes, no charge
    Runs as
    A self-serve assessment
    What you receive
    Fifteen questions on lineage, quality, ownership, access and platform, with an instant result and no email asked for. Worth sending to every portfolio CFO before the next board meeting.
    Send it to a portfolio CFO

After either of the first two: a first build at one portfolio company, measured against a number already in its P&L, then a decision to take it across the portfolio, change it or stop.

AI & Agent Pilot · Data Platform Accelerator

Q&A log: what deal teams, operating partners and CFOs ask

Each question as it would sit in the data room’s Q&A log, with its answer.

Can you work inside the deal timetable?

Yes. Data diligence runs alongside the financial and commercial work, in the data room or in the company’s systems where access is given, and reports in the format the deal team already uses. Two weeks covers one company’s revenue, customer and margin data.

We have no system access before signing. Is it still worth doing?

Yes, with less depth. From the data room’s exports we can test whether the numbers in the CIM can be rebuilt and show where they cannot; access after signing confirms the rest in the first weeks.

How do you keep AI value in the plan honest?

Every initiative is tied to a line already in the P&L, with a baseline and a control group, and its run-rate counts only once it has been measured in production. That is also how a buyer’s quality-of-earnings review tends to look at it.

Can one approach work across very different portfolio companies?

The definitions and the reporting feed are shared; the data work is sized to each company. A distribution business and a software business get different builds, and report the same KPIs to the fund.

Who owns the data and the tools when the company is sold?

The portfolio company. We build in its accounts and under its contracts, so everything transfers with the business and nothing depends on the fund or on us.

Will deal data or investor data train anyone’s model?

No. We work inside your environment or the company’s, under the data room’s confidentiality terms, and the AI services are set up so that providers neither keep nor train on what is sent.

Do you work for the fund or for the portfolio company?

Either, and often both: diligence and portfolio reporting for the fund, the build for the company. The engagement letter says which, so it is clear who the work is for and who pays for it.

Can you help get a company ready to sell?

Yes. The KPIs, cohorts and AI evidence a buyer asks for are built from the systems well before the sale, so the sell-side data room comes together in days and the numbers survive the buyer’s diligence.

Bring the next deal, or the company furthest behind its plan.

A CIM whose numbers you want tested, a value creation plan with AI in it, a portfolio that reports twelve ways, or an exit eighteen months away. Thirty minutes with whoever would lead the engagement, on the first thing to check.

Book 30 minutes on your deal

AI in the plan, not in the P&L? Thirty minutes on a deal, a portfolio company or an exit.

Book a call on your deal