Buy Certainty First: The Fixed-Price Evidence Product That Ends the Bespoke SOW

Estimation and negotiation happen at once over unpaid speculative labour — so neither party trusts the number. Separate the purchases: sell a fixed-price evidence product first, compile the SOW as a receipt, and measure the sales system for the first time.

Scott Farrell · LeverageAI · Commercial protocol for professional services

Ask a mid-sized consultancy why proposals cost so much, win so rarely, and still lose margin when they win, and you will get a story about competition, procurement, or the economy. Those forces are real. They are not the design flaw.

The design flaw is the entry transaction itself. Estimation and negotiation happen simultaneously over unpaid speculative labour. The client protects its budget. The firm protects its margin. Both sides know the statement of work is partly fiction. Delivery then discovers the estate the proposal pretended to understand. Change requests finish what the SOW began.

This piece is about a different commercial protocol: buy certainty first. The first paid unit is not the beginning of implementation. It is a bounded reduction in uncertainty — a real option on transformation — after which budget discussion happens honestly over evidence-backed options. The SOW stops being authored and starts being compiled. Fixed price becomes rational because unknowns are typed deliverables, not heroic assumptions. And for the first time the sales system becomes measurable as a system, not as a personality contest.

The specimen is one production product — the Data Readiness Review, built on FDE BI — and I will treat n=1 as n=1. The protocol is demonstrated, not surveyed across a portfolio. The argument travels further than data consulting; the proof burden does not get inflated to match the ambition.

The senior SOW author is the join algorithm

I have lived the bottleneck. In services-only firms, the bespoke statement of work was so slow and unreliable that I had to write the hard ones myself. Multiple seniors, multiple days, unpaid. The definition of done was wild. One answer could land at two hundred thousand dollars; another, same brief, six hundred. It depended who did it and on what day of the week. Looking over either number and declaring it "correct" was almost impossible — there was no substrate to test it against.

That is not a talent problem. It is an architecture problem.

Today a senior consultant writing an SOW manually joins:

CLIENT REQUEST
+ ACCOUNT HISTORY
+ FIRM CAPABILITIES
+ AVAILABLE PEOPLE
+ PRIOR ARCHITECTURES
+ TECHNOLOGY CONSTRAINTS
+ COMMERCIAL RISK
+ DELIVERY ASSUMPTIONS
+ PERSONAL EXPERIENCE
        ↓
SENIOR PERSON MANUALLY COMPUTES
        ↓
SOW

The institution does not reliably hold that joined context. So a person reconstructs it from emails, meetings, colleagues, old proposals and instinct. The senior SOW author is the join algorithm. Both the $200k and the $600k answers can be locally rational because each author assembled a different world.

AI will not attack this primarily by writing prettier proposal prose. That is the least interesting attack. The interesting attack is an alternative firm that retains prior engagement knowledge, measures the estate before pricing, exposes assumptions, reuses tested work packages, and generates a defensible commercial decision — of which the SOW is only the receipt.

The native output is not a better-written SOW. It is a defensible commercial decision, from which the SOW becomes a generated receipt.

Win-rate economics and wrong friction

Nobody pays you to write the statement of work. Yet it is hard, senior, and expensive. You do not win them all. At a ten-percent win rate — an economic shape, not a claim about every firm's average — the winning job effectively absorbs the pre-sales effort of ten attempts. Then the won project can lose more money through omitted assumptions, free accommodation, scope creep, handoffs and change negotiation.

The firm calls this business development. Economically, a large portion is unpriced cognition spent reconstructing the same commercial object from scratch.

The Friction Attack Surface taxonomy is the right sorting hat here. Do not re-litigate it; use it.¹

• Necessary friction: actual client decisions, authority, contractual risk, unusual architecture, accountability, genuine new intent.
• Accidental friction: searching old projects, finding the specialist, copying boilerplate, reconciling versions, rebuilding pricing assumptions by hand.
• Wrong friction: every engagement made broad and bespoke because the commercial model sells teams and days rather than bounded outcomes.

Protect the first. Compile away the second. The third may require changing the business model itself. The budget–proposal dance sits largely in the third category: unpaid discovery because the market convention says proposals are free; soft assumptions so a price can be submitted; underquoting to win and recovering margin through changes; clients comparing proposals built on different implied realities.

That does not exist because the client's outcome requires it. It exists because traditional consulting sells uncertain projects through speculative labour.

The conventional transaction is badly designed

Client has an incompletely understood problem
        ↓
Consultancy performs substantial unpaid investigation
        ↓
Both parties negotiate price while neither knows the true scope
        ↓
Client protects its budget · Consultancy protects its margin
        ↓
A strategically ambiguous SOW is agreed
        ↓
Reality is discovered during delivery
        ↓
Change requests, margin erosion, arguments about assumptions

The replacement protocol is not "write proposals faster." It is a different market transaction:

Client buys a bounded reduction in uncertainty
        ↓
Evidence is collected and compiled
        ↓
Current state, gaps and unknowns become explicit
        ↓
Humans decide what matters
        ↓
Implementation options are priced against observed reality

You have separated the purchase of certainty from the purchase of implementation.

That sentence is the commercial heart of the protocol.

Separate the purchases — offer a mirror, not a verdict

The budget discussion is not illegitimate. Clients have constraints; firms need viable engagements. The problem is entanglement of two activities:

1. Estimation: what work exists, what is uncertain, what will it take?
2. Negotiation: which outcome, risk allocation and price will the parties accept?

When those happen together, neither party trusts the number. The client thinks: you asked my budget so you could quote exactly that amount. The firm thinks: you are withholding information while expecting us to assume delivery risk.

The fixed-price certainty product disentangles them. The first transaction says: for a known amount, we will establish enough shared evidence to make the second commercial decision intelligently. Afterward, budget can be discussed honestly:

• Option A costs this and delivers this.
• Option B fits the available budget and consciously defers these items.
• Option C is not currently justified.
• Doing nothing remains a legitimate disposition.

The budget no longer secretly determines the diagnosis. It chooses among evidence-backed options.

Economically this is a real option on transformation: the right, not the obligation, to invest intelligently later. For a bounded commitment the client receives a map, gaps, unknowns, required decisions, options and a defensible scope. They can proceed, defer, reduce, use another supplier, or decide the investment is not justified. That optionality is why it feels safe.

The firm receives payment for work previously absorbed as pre-sales, evidence-based orientation to the account, a better-qualified implementation opportunity, and a credible basis for pricing. That is earned incumbency through superior orientation — not lock-in through obstruction.

The SOW is compiled, not authored

The inversion is simple to state and expensive to implement.

Old path: vague request → meetings → opinions → estimated architecture → negotiated number → SOW → discover reality during delivery.

New path: observed evidence → declared requirement → mapped gap → explicit questions → human decisions → bounded work packages → SOW.

Correctness becomes reviewable against visible things: which findings entered scope; which a person accepted; which assumptions remain; what evidence supports each work item; what is excluded; what the client must supply; what acceptance test defines completion; which uncertainties require reserve rather than pretence of knowledge.

A reviewer may still disagree with the commercial number. They are no longer reviewing an opaque lump of prose. They can see the compiled basis.

This is where the book stops and Proof-Carrying Transformation begins. Advice got cheap; verification did not.² The delivery-side chain — evidence through decision, build and learning — is PCT's territory. The certainty product is the commercial airlock into that chain. Do not break the spine by reconstructing reality again at kick-off.

Fixed price without heroics: typed uncertainty and the pricing envelope

Fixed price for an uncertain activity is a risk transfer from client to supplier. A weak supplier cannot safely do that. It overprices, underprices, neuters the assessment, or recovers through change requests.

The breakthrough is not bravery. It is typed uncertainty.

Valid terminal states include: directly mapped; partially mapped; not observed; insufficient evidence; inaccessible within the audit boundary; unsupported source type; ambiguous — consultant decision required; excluded from this phase. Those are not failures of the product. They are part of the product.

Unknowns become typed deliverables instead of unbounded consulting labour.

The promise is not "we will completely understand and solve every item in your estate." The promise is: we will produce a defensible inventory of what is present, what maps, what does not, what remains unknown, what decisions are required, and what should enter the next scope.

The operating envelope is machine-measured product configuration, not estimation:

• automated preflight census (workbooks, sheets, tables, formulas, connections, queries, semantic models, measures, requirements, early ambiguity rates);
• source and workbook volume bands;
• included number of material findings or review cases;
• explicit unsupported-source treatment;
• a Flex Reserve for unusual access or exception density;
• contractual boundary that "not observed" does not mean "does not exist".

Worked pricing envelope (shape, not invented rates)

A concrete mechanical walk — the numbers below are structural placeholders for band design, not a published price list and not portfolio statistics:

1. Census metrics: count sources, workbooks, sheets, model objects, declared requirements; flag access gaps and early ambiguity.
2. Band assignment: map census into Band S / M / L by thresholds you publish (for example: workbook count and source-system count). Band chooses fixed fee and calendar timebox.
3. Included finding counts: each band includes N material findings through human disposition. Beyond N is not free heroics — it is either a higher band or Flex Reserve drawdown.
4. Flex Reserve trigger: exception classes (inaccessible systems, unusual source types, security review delays, finding density above band) consume reserve units. Exhaustion reopens commercial conversation with evidence, not surprise.
5. Boundary language: not observed / insufficient evidence remain deliverable states; they do not silently become "absent from the estate."

That is configuration of a product. It is not a senior person inventing a number on a Thursday.

The machinery that makes this deliverable — deterministic sensors, privilege separation, human disposition of model findings — is the substrate developed in the sibling books on AI-constituted services and separation of powers for cognition.^3,4 One sentence is enough here: without that stack, the fixed-price promise is reckless or padded. With it, predictability is a premium attribute.

Commercial legibility: Red Beads, two lanes, real KPIs

The most valuable organisational insight is not the fixed price. It is commercial legibility.

The old RFQ funnel is not one process. It is unrelated processes masquerading as a funnel — different offerings, discovery depths, buyer maturity, senior effort, discounting, hidden assumptions, competitors, delivery feasibility — summarised into one "win rate" that barely measures sales effectiveness.

A stable unit makes the business learnable:

Defined eligibility
→ named buyer problem
→ standard commercial promise
→ fixed price or price band
→ defined timebox
→ repeatable evidence process
→ standard decision outputs
→ observable downstream options

Productisation does not merely make the service easier to buy. It makes the business capable of learning from selling it.

Deming's Red Bead Experiment is the right reference and must be used honestly. Beginning in the early 1980s, Deming used the experiment to illustrate the fallacy of ranking people for performance when their results are constrained by the system they work in. A willing worker wants to do a good job; success is limited by the system; sustainable improvement comes when management improves the system.⁵

Consulting sales is a Red Bead environment when management attributes mixed outcomes to individuals. Raw wins and losses are affected by account quality, incumbency, relationship history, timing, procurement rules, competitive price-cutting, proposal quality, technical credibility, delivery capacity, product attractiveness, and how much risk the buyer is asked to assume. A salesperson can raise win rate by discounting and damage margin. Another can lose a poorly qualified opportunity that should never have entered the funnel.

What the readiness product does is put more of the causal chain under control: eligible accounts, a concrete product, price integrity, decision advance, objection capture, exception routing. Management can separate:

1. Account conditions
2. Sales execution
3. Offer performance
4. Delivery performance
5. Downstream performance

That separation is what makes KPIs meaningful.

Run two visibly separate commercial lanes

Do not blend the readiness offer back into general proposal statistics.

Legacy RFQ lane — track: unpaid sales and technical hours; time to submission; win rate; discounting; forecast margin; actual margin; estimate variance; change-request frequency; reason for loss (with uncertainty acknowledged).

Productised readiness lane — track:

• Front door: eligible accounts identified; percentage offered the product; conversation-to-purchase conversion; days from offer to decision; salesperson and technical hours before purchase; price discounting.
• Review economics: delivery cost and gross margin by input band; sources/workbooks/findings processed; consultant review effort per material finding; exception and escalation rate; cycle time.
• Downstream quality: review-to-build conversion; estimated versus actual build effort and margin; change-request frequency and value; scope items later found unsupported or omitted; time from review completion to implementation decision.
• Compounding: reusable sensors or tests created; exception classes that no longer require central escalation; engagement-two consultant independence; reduction in cost or review effort across successive assessments.

The comparison is not which lane has the higher close rate. It is which lane produces more total economic value per eligible opportunity — with less client effort, less speculative labour, better downstream margin and more reusable learning.

Internal deployment work already points the same direction: measure the funnel, not the vibes.⁶

The client–consultancy friction map

The unit of analysis is not "where can we use AI?" It is every boundary touch between firm and client where either party performs translation, reconciliation, uncertainty management or coordination the outcome does not require.

Apply a Human Touch Audit to pre-sales: for every human appearance, ask whether judgment, new intent, authority or exceptional consequence was required. A senior architect re-reading five old proposals to discover whether the firm has done similar work is exposed friction. A sponsor accepting residual risk is authority — preserve it.

Specimen: Data Readiness Review (n=1)

The Data Readiness Review is the working instance of the protocol, not a portfolio proof. It is an evidence-backed consulting transaction compiler: observed estate → declared target → interpreted comparison → human decisions → readiness position → scoped next engagement.

One synthetic report shape (from the specimen build): distinctions such as found / partially supported / not found among scope-bearing findings, with all findings remaining unreviewed model positions until a human disposes them. The scope carries uncertainty forward — explicit exclusions, client responsibilities, acceptance criteria that can fail — rather than erasing the uncertainty that produced it.

A legitimate "not ready / do not build" outcome is commercial success when the refusal is real: the client avoids a bad investment; the firm demonstrates independence; expensive delivery capacity is not consumed by a poor project. If the review cannot honestly recommend non-proceeding, it is disguised lead generation and the trust advantage disappears.

The financial case is broader than review fee × volume:

Readiness-review revenue
+ avoided unpaid proposal effort
+ shorter time to paid engagement
+ greater conversion into implementation
+ improved implementation gross margin
+ fewer unpriced scope surprises
+ lower change-control friction
+ better account expansion
+ reusable delivery learning

Some terms may dominate others. Do not invent which ones; measure them in your two lanes.

What this is not

This is not proposal automation for answering client RFPs — that is the Intelligent RFP problem, the opposite direction of the transaction. It is not Marketplace of One outbound (proactive bespoke proposals as the opening move), though it shares the conviction that the sales artefact should prove the method.⁷ It is not the full delivery engagement chain after the sale — that is Proof-Carrying Transformation. It is not the sensor architecture or the AI-constituted service taxonomy — those are sibling organs. Vendor-side moat design is a later problem; this book owns the entry protocol and its measurement.

Monday morning

If you run a professional services firm with expensive pre-sales, do this:

1. Map ten boundary touchpoints. Classify each necessary / accidental / wrong.
2. Find where a senior person is the join algorithm.
3. Name the first bounded uncertainty that can become a paid evidence product.
4. Define typed terminal states and a pricing envelope (census → band → included findings → Flex Reserve).
5. Separate the lane: do not blend metrics with legacy RFQs.
6. Instrument front-door, review economics, downstream quality and compounding.
7. Allow do-not-build as success.
8. Hand the evidence spine into delivery without reconstruction — then stop this book's job and start PCT's.

The nebulous consulting engagement used to be an advantage. Under cheaper cognition it becomes friction rent — hard to explain, hard to buy, hard to scope, hard to price, hard to govern and hard to know when finished. The answer is not a cheaper quote before you understand the estate. The answer is a product for understanding it before you ask anyone to commit.

Buy certainty before you buy the transformation.