# Portfolio AI Ops — Pricing Three tiers. Fixed fee, fixed term, no per-seat component and no usage billing. Every number below is the actual number quoted, not a starting point for negotiation. --- ## 1. The tiers ### Diagnostic — $18,000 one-time Three weeks. No retainer, no term commitment. - Up to **3 portfolio companies** instrumented - Retroactive scoring of the **last 4 quarters** of AI claims already made in the portfolio - Baseline calibration report: what the fund's existing AI claims would have scored - A sized, underwritten **forecast slate** for the next cohort, agreed with each company CFO - Deployment map: what is worth building, what is not, and in what order **Credits in full against the Deployment setup fee.** **Guarantee:** if it does not surface at least $500K of measurable annualised upside across the three companies, it is free. --- ### Deployment — $45,000 setup + $12,000/month Six-month minimum term. Most funds start here after the diagnostic. - Up to **4 portfolio companies**, systems live in production - Full forecast ledger, hosted, LP-shareable - **Quarterly** resolution cycle with a computed IC memo - Direct access to the engineer who built the systems, not an account manager - **+$2,500/month per additional portfolio company** Year one, 4 companies: **$189,000** --- ### Portfolio Standard — $60,000 setup + $22,000/month Twelve-month term. The forecast ledger becomes the fund's operating standard. - Up to **10 portfolio companies** - New platform companies instrumented **within 30 days of close** - LP-ready quarterly reporting pack - Diligence support on AI claims made by targets in live deals - Operating-partner training on forecast underwriting (the fund keeps the discipline even if I leave) - **+$2,000/month per company beyond 10** Year one, 10 companies: **$324,000** — $32,400 per company per year. --- ## 2. What it replaces, in dollars | Alternative | Year-one cost | What you actually get | Source of estimate | |---|---|---|---| | **MBB / Big-4 digital diligence** | $500K – $2M | A recommendation deck. No deployment, no measurement, no one carries the forecast | Published MBB rate cards: $180K–$250K per consultant-month, typical 3–5 month team | | **Director of Portfolio Technology** (hire) | $280K loaded ($195K base + 30% benefits/tax + recruiter fee) + 6–9 months to fill | Real capability, eventually. Year one is orientation. Single point of failure who can leave | Heidrick & Struggles / BLS comp data for the title, LMM funds | | **Three AI point vendors × 4 companies** | $180K – $400K/yr | Seats and usage dashboards. EBITDA attribution is left to you | Typical mid-market vertical AI SaaS: $1.5K–$8K per company per month | | **One unmeasured internal pilot** | $150K – $250K + ~$85K of internal time | An anecdote. No defensible attribution | Observed range across LMM portfolio AI pilots, 3–4 month duration | | **Do nothing** | $0 | Defensible for one more cycle. Not two | — | | **→ Portfolio AI Ops, Deployment** | **$189K** | Systems in production in 4 companies, plus a scored, falsifiable ledger | This document | The honest framing: this is **not** cheaper than doing nothing, and it is **more expensive** than a single point vendor. It is cheaper than one MBB engagement, cheaper than one bad hire, and roughly the price of the pilot you would have run anyway to learn less. --- ## 3. Buyer payback math Deployment tier, four companies, deliberately conservative. ``` COST Setup $45,000 Retainer, 12 months × $12,000 $144,000 Total year-one cost $189,000 VALUE (reference ledger runs 7.3x; this models 3.0x) Realised value, 12 months $567,000 Net EBITDA effect $378,000 RETURN Payback period $189,000 / ($567,000 / 12) 4.0 months Year-one ROI on fee 200% Enterprise value at 9x exit multiple $3,402,000 Return on fee, EV basis 18 : 1 ``` **Sensitivity — what if it underperforms?** | Realised multiple | Realised value | Net effect | Verdict | |---|---|---|---| | 7.3x (reference ledger) | $1,380,000 | +$1,191,000 | Exceptional | | 3.0x (base case above) | $567,000 | +$378,000 | Strong | | 1.5x | $284,000 | +$95,000 | Marginal but positive | | **1.0x (break-even)** | **$189,000** | **$0** | The programme paid for itself and you kept the ledger | | 0.5x | $95,000 | −$94,000 | Loss. Kill at the 6-month term boundary | Break-even sits at a **1.0x** realised multiple. The reference ledger runs 7.3x. The gap between those two numbers is the entire margin of safety in the offer, and it is why the six-month minimum term is short enough to be honest. --- ## 4. Margin math (internal) Solo operator. Inference runs on owned GPU hardware — a 5-node local fleet totalling ~293GB VRAM — so marginal inference cost is electricity, not API spend. That is the structural cost advantage and it is why these prices work solo. ### COGS per account per month — Deployment tier, 4 companies | Component | Detail | Monthly cost | |---|---|---| | Delivery time | 26 hrs/mo at a $145/hr implied opportunity rate | $3,770 | | Inference | Local fleet, amortised hardware + power. ~$0.00 marginal, $310 amortised | $310 | | Hosting / ledger infra | Ledger hosting, backups, per-company isolation | $180 | | Third-party data | Apollo, enrichment, monitoring | $120 | | Software / tooling | Allocated | $95 | | **Total COGS** | | **$4,475** | ``` Revenue per account per month $12,000 COGS per account per month $4,475 Gross profit per account per month $7,525 Gross margin 62.7% ``` Including the amortised setup fee over the first 12 months ($45,000 / 12 = $3,750/mo against roughly 40 hours of front-loaded onboarding, ~$1,930/mo amortised cost), blended year-one gross margin on a Deployment account is **≈66%**. ### Capacity ceiling | Constraint | Number | |---|---| | Delivery hours available per month (solo, 60% billable of a 55-hr week) | ~143 hrs | | Hours per Deployment account per month | 26 hrs | | **Maximum concurrent Deployment accounts** | **5** | | Revenue at 5 accounts | $60,000/mo = **$720K/yr** | | Gross profit at 5 accounts | ~$451K/yr | **Hire trigger: account #4.** At four concurrent accounts (104 delivery hours) there is no slack left for sales, and the pipeline dies exactly when it is most valuable. The first hire is a deployment engineer at ~$140K loaded, which moves the ceiling to roughly 11 accounts and drops gross margin to ~54% before it recovers on the next tier of scale. **The real constraint is not GPU capacity, it is Mike-hours.** The fleet could serve 50 accounts. One person cannot underwrite 50 forecast slates. --- ## 5. Expansion path ``` Diagnostic $18K one-time | credits 100% against setup; ~55% convert within 60 days v Deployment $45K + $12K/mo (4 companies) | +$2,500/mo per extra company | natural expansion: companies 5, 6, 7 added one per quarter v Deployment at 7 companies $19,500/mo | at 8+ companies, Portfolio Standard is cheaper for the buyer — say so v Portfolio Standard $60K + $22K/mo (10 companies) | +$2,000/mo beyond 10 v Fund-level standard: every new platform instrumented within 30 days of close ``` **Per-company add-on:** $2,500/mo on Deployment, $2,000/mo on Portfolio Standard. **Tell the buyer when to upgrade.** Deployment costs $12,000 + $2,500 per company beyond four, so it crosses Portfolio Standard's flat $22,000/mo at exactly **8 companies** and is more expensive at 9 or more: | Companies | Deployment /mo | Portfolio Standard /mo | Cheaper | |---|---|---|---| | 6 | $17,000 | $22,000 | Deployment | | 7 | $19,500 | $22,000 | Deployment | | 8 | $22,000 | $22,000 | Equal — recommend the upgrade here for the extra scope | | 9 | $24,500 | $22,000 | **Portfolio Standard** | | 10 | $27,000 | $22,000 | **Portfolio Standard** | Volunteering that math at company seven, before they find it themselves, is worth more than the margin it gives up. It is the moment the relationship stops being vendor and starts being advisor. **The real expansion is not more companies. It is the next fund.** An operating partner who has run the ledger for a year takes it with them, and instrumenting a new platform at close is the highest-margin work in the model — the forecast slate is written before anyone has formed bad habits. --- ## 6. Why we price this way **Fixed fee, not per-seat.** Per-seat pricing punishes exactly the behaviour the product exists to encourage. If instrumenting a fifth company costs more per user, the operating partner instruments fewer companies, the ledger gets smaller, n falls, and the calibration statistics stop meaning anything. The product only works at portfolio scale, so it has to be priced at portfolio scale. **The diagnostic is priced to be paid for, not to be free.** A free assessment gets treated exactly as seriously as it cost: the CFO does not clear time, the data access request sits for three weeks, and the document is never actioned. $18K is small enough to clear without a committee and large enough that someone senior shows up. **Setup fee is real, not a discounting instrument.** Weeks one to four are the heaviest work in the entire engagement — data access, baselining, and underwriting the first forecast slate with each CFO. Charging for it filters out funds that were never going to grant data access, which is the single largest cause of failed engagements in this category. **Six-month minimum on Deployment, not twelve.** Two quarterly resolution cycles is the shortest honest test of the thesis — one cycle is noise. Asking for twelve months would be asking the buyer to commit before the product has produced a single scored number, which contradicts everything the product claims to stand for. **No success fee, no share of savings.** Tempting, and wrong. A success fee gives the person who computes the score a direct financial stake in the score, which destroys the only asset being sold. The measurement has to be boring and disinterested or it is worth nothing.