Sell the Compression, Not the Components
If your real capability spans business intent through governed production, a capability catalogue makes that truth sound like a lie. Design one impossible compression—and a ladder of claims the buyer can climb.
- Unusually broad FDE span fails as marketing when you itemise every discipline. Lists multiply disbelief even when each line is accurate.
- Sell one impossible compression: a company-specific problem that moves through research, decision, architecture, security, build, tests, and write-back without handoff loss.
- Use a positioning ladder—credible public claim, unusual demonstrated claim, extraordinary underlying claim discovered by the buyer—and prove only four modest visible claims.
Here is the marketing problem nobody puts on the capability slide.
You can actually take a messy executive concern, frame it, connect it to architecture and governance, build something that runs, show tests, and leave the next engagement smarter than this one. That closed loop is real. It is also almost impossible to say out loud without sounding like a brochure written by a committee of your own egos.
I have sat with the sentence that tries to list it all—strategy, business case, ARB pack, cyber view, implementation, deployment, evaluation, learning—and watched the room’s eyes glaze into polite disbelief. The more complete the list, the less believable each item becomes. As one hard night of self-assessment put it: everything sounds like hyperbole. It can be accurate and still sound ridiculous.
That is not a personal-branding problem. It is a proof design problem. Buyers have been watching GenAI pilots fail to become production impact. MIT’s NANDA initiative’s 2025 State of AI in Business study—the “GenAI Divide” report—found that, in its dataset, roughly 95% of organisations were getting no measurable P&L return from GenAI investment, with only a thin edge of integrated pilots extracting real value while the vast majority remained stuck without measurable P&L impact.1 In that climate, a longer capability catalogue does not read as seniority. It reads as another promise stack.
So stop selling the components. Sell the compression.
The unit buyers can believe
Forward-deployed engineering, as the market now names it, is not a feature matrix. OpenAI’s deployment arm defines FDE as bringing AI into production for complex real-world use cases: work a specific customer problem, validate impact, then generalise patterns that scale—systems that work in practice, not only in theory.2 The commercial unit is a joined intervention, not a headcount of skills.
Your headline is not: “I possess an extraordinary number of capabilities.”
It is closer to: I collapse the distance between an executive problem and a governed production system—without losing the reasoning between them.
Buyers already understand the fragmented chain: strategy consultant → analyst → architect → cyber → engineering → test → deploy → ops. What they do not believe, from a list, is that one spine can hold that chain with fidelity. What they can believe—if you show it—is one problem moving through those stages without the usual telephone-game losses.
That is the impossible compression: one outcome that normally requires a team, multiple handoffs, and months, shown as a single joined pass with inspectable artefacts. You are not claiming to be the world’s best specialist in every discipline. You are claiming to run a delivery system that coordinates those disciplines without discarding the “why” at every boundary.
RAND’s interviews with AI practitioners still land the organisational half of this: AI projects fail roughly twice as often as non-AI IT work (by common estimates), and the leading failure pattern is misunderstanding or miscommunicating the problem to be solved—optimising the wrong metrics, chasing technology for its own sake, under-investing in the path to production.3 Their practical line is blunt: focus on the problem, not the technology.4 A demo that starts with your stack is the wrong object. A demo that starts with their discontinuity is the right one.
Why catalogues make truth sound false
Capability catalogues fail for a mechanical reason: each added line is an additional claim that must be believed in parallel. Belief does not work that way for extraordinary breadth. Trust accumulates when a buyer watches coherence across stages—when the business case still matches the architecture, the architecture still matches the threat model, and the code still matches the acceptance criteria they saw ten minutes earlier.
List mode invites procurement theatre: “Which of these boxes do you tick?” Compression mode invites a different question: “Did that same problem stay joined?”
There is a second reason catalogues fail now. Pretty decks got cheap. Advice without implementation path is under commercial and cultural pressure. In Australia, KPMG’s FY25 impact release put consulting revenue down 18% on reduced government use of consultants and a broader slowdown—while the firm described rebalancing traditional consulting toward technology transformation and AI.5 That is not “consulting is dead.” It is a bifurcation signal: ceremonial recommendation is easier to discount; applied transformation still gets the reallocation story.
If your anti-pattern is management consulting that leaves a set of slides with no rubber on the road—no provenance, no governance inputs, no path through architecture review, no test against the real world—then your marketing must not recreate the same shape with better adjectives. “I’ll give you receipts” only works if the first contact already feels like receipts, not like a feature brochure.
The positioning ladder
UX designers solved a version of this problem decades ago. Progressive disclosure shows users only the few options that matter first, and reveals advanced complexity on request—so systems stay learnable without lying about depth.6 Credibility marketing for broad operators needs the same physics.
Call it a positioning ladder. Three rungs. Climb in order. Never open on the top rung.
| Rung | Claim | Job |
|---|---|---|
| Public | I lead complex AI initiatives from discovery through governed production. | Comparable, hireable, not grandiose. Safe on a website and in a cold intro. |
| Demonstrated | I can produce strategy, business case, stakeholder designs, and working implementation through one integrated delivery environment—on your problem. | Unusual, but visibly provable in a short proof path. |
| Underlying | I have compiled judgement, code, conversations, and frameworks into a portable operating kernel that compounds across engagements. | Extraordinary. This is the conclusion the buyer reaches—not your opening boast. |
The ladder is not false modesty. It is belief design. The public claim is true and modest. The demonstrated claim is where your storyboard lives. The underlying claim is what makes the compression possible—and if you lead with it, you sound like you are selling a religion. If you end with it, after the buyer has watched one problem stay coherent, they invent the extraordinary claim themselves. That is stronger than any slide that says “28 years of experience, fully indexed.”
Adjacent work already named the proposal version of this: meta-credibility—the way you sold them is the way you’ll serve them; the pre-engagement artefact is the first deliverable, not a costume. Company-specific proposal compilation belongs to that lane. This piece is the demo sibling: the same property applied to progressive proof on camera and in the room.
Four claims—only four
Do not demonstrate “everything.” Demonstrate four things the buyer can verify without trusting your biography:
- I understand unusually quickly. Company-specific evidence shows up early. Not a generic industry deck.
- I connect business, technology, and governance without handoff loss. The same problem appears in executive language and in architecture/security language without rewriting the truth between them.
- I can turn the analysis into something working. Not a slide about a build. A vertical slice, a test, a deployment path—something that has contact with reality.
- The work leaves the delivery system smarter than it began. Write-back is visible: a learning, an eval case, a rejected path filed, a kernel update that is allowed to compound.
Once those four are believed, breadth becomes plausible. Without them, the breadth list is inflation. Notice what is missing: no claim to be the best cryptographer, the best change manager, or the only person who can hire five specialists. The visible claim set is deliberately modest. The compression does the heavy lifting.
The 90-second storyboard
This is the primary artefact. Not a five-minute autobiography. Not “here are my frameworks.” A short proof path you can script, film, or walk live.
Storyboard: one impossible compression
One live receipt, in plain language
Suppose the buyer’s discontinuity is: “We cannot tell which AI use cases should enter production, so we fund pilots that never join.” A compression demo does not open with your methodology names. It opens with their ambiguity, then shows a single spine:
- Recommendation exhibit: a one-page decision pack—three options, rejected paths with reasons, a recommended lane, and what would falsify it.
- Architecture exhibit: the same problem restated as system boundaries, data movement, and where the model is allowed to act versus where deterministic code or a human gate owns the decision.
- Implementation exhibit: a thin vertical slice that implements that boundary—not a greenfield rewrite of their estate.
- Test exhibit: eval cases or acceptance checks that map back to the decision pack’s success criteria, including at least one unhappy path.
That chain is the receipt. The labels can be rough. The join cannot be. If the architecture page cannot point at the recommendation it realises, or the test cannot point at the architecture it enforces, you are still performing catalogue theatre with better props.
Enterprise sales already distinguishes lightweight guided demos from full proofs of value that use the customer’s data and processes.7 Treat the storyboard as progressive proof on a spectrum—not as an unpaid six-week pilot by default, and not as a product tour of your favourite side project. The test is simple: after ninety seconds, can the viewer restate the four claims without seeing a capability list? If they instead recite your tool names, you sold components.
Role-specific cuts, one ground truth
The deep system underneath stays fixed. The first frame changes by who is in the room. That is radial translation, not custom fiction: same joined ground truth, first-generation views cut for the destination register so nobody is more than one hop from the truth.
| Buyer | Open on | Still prove (somewhere in the cut) |
|---|---|---|
| CIO / CTO | Cross-system operating picture, architecture alternatives, integration and debt, governed delivery path | All four claims; one live receipt into build/test |
| CISO / risk | Threat surfaces, data movement, authority boundaries, deterministic controls, audit receipts | Same ground truth; code and business case still available on request |
| CEO / P&L owner | Business discontinuity, strategic alternatives, investment case, one concrete operating intervention | Path to working system and evaluation—not a pure strategy monologue |
| Consulting firm leader | Client-specific pack, multiple audience registers from one spine, how engagement learning becomes reusable firm capacity | That the method is experienced, not only described |
If you produce four unrelated decks that drift into four different “truths,” you have recreated the telephone game you claim to fix. The cut is a lens. The spine is not negotiable.
The engagement continues the ladder
Marketing is not separate theatre. The proposal (when you use one) is evidence of how you will engage. The engagement is evidence of how the system works. The deployed result is evidence the engagement worked. The next engagement is evidence that learning compounded.
That last step is the quiet moat. Plenty of people can deliver one impressive project through heroics. The claim worth believing is that each project leaves machinery better able to perform the next—without trapping the learning only in one head. In operator language: the engagement is the evidence of how it works. You are not merely bringing a “wiki of everything” into a room. You are conducting the work through a compiled field, and the work emits the receipts that validate—and improve—the field.
That continuity is also how you avoid the other failure mode: a dazzling pre-sales demo that the delivery team cannot re-run. If sales used a different spine from delivery, you bought meta-credibility for a week and spent it on day one of the engagement. Progressive proof only compounds when the same ladder keeps going after the contract signature—discovery package, decision package, architecture and governance, build with tests, production observation, learning write-back.
So the commercial packaging should make breadth feel necessary, not boastful. A named intervention—discover, decide, build, deploy-and-learn—explains why the disciplines appear. You are not listing talents. You are listing what the job requires to stay joined. Pricing against salary bands recreates the wrong comparison; that is a different article. Here the point is only this: sell the completed compression, not the org-chart silhouette of the person who runs it.
What this is not
- Not a clever off-problem demo. If the buyer classifies it as “your toy,” you lost. Compile the demonstration for them.
- Not a capability inventory with better typography. Frameworks can be present as ambient conditioning; they are not the exhibit.
- Not a claim that you never use specialists. Compression is coordination with fidelity, not omniscience cosplay.
- Not a substitute for diligence. Progressive proof opens belief; governance reviews, security, and production ownership still earn the later rungs.
- Not a Practice OS sales pitch. How you industrialise a firm’s bench is another conversation. This one is how one operator (or a small pod) becomes believable.
Script this week
- Pick one real buyer and one real discontinuity (sanitise if you must publish).
- Write the three ladder claims on a single page. Delete any public sentence that belongs on the top rung.
- Storyboard ninety seconds using the beats above. Time it.
- Attach one live receipt: recommendation → architecture decision → implementation artefact → test or eval result.
- Cut a CIO-first and CISO-first opening from the same spine.
- Show it to two skeptical humans. Ask them to restate the four claims without prompting. If they recite your tool list instead, you still sold components.
Major labs already define forward-deployed engineering as the motion that takes AI into production for complex, real-world use cases.2 The title will get washed. What will not wash is a buyer who has watched one of their own problems stay coherent from intent to evidence—and then realised, without being told, that a larger machine must exist underneath.
You do not need the squeaky bionic sound effect. You need progressive proof.
Sell the compression. Let them discover the components.
Next step: Script one ninety-second cut for one buyer this week. If you want the deeper treatment of the compiled career behind the compression, see Compounded Execution Capital; for engagement evidence packages that keep verification risk with the deliverer, see Proof-Carrying Transformation. Both live on leverageai.com.au under static articles 165 and 164.
References
- MIT NANDA Initiative. “The GenAI Divide: State of AI in Business 2025.” — “Despite $30–40 billion in enterprise investment into GenAI, this report uncovers 95% of organizations are getting zero return. Just 5% of integrated AI pilots are extracting millions in value, while the vast majority remain stuck with no measurable P&L impact.” https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf
- OpenAI Deployment Company. “What is forward deployed engineering (FDE)?” — “Forward deployed engineering (FDE) is how OpenAI brings AI into production for complex, real-world use cases… solve a specific problem, validate impact, and then identify patterns that can scale.” https://deploy.co/en-US
- RAND Corporation. “The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed.” — “By some estimates, more than 80 percent of AI projects fail—twice the rate of failure for information technology projects that do not involve AI.” https://www.rand.org/pubs/research_reports/RRA2680-1.html
- RAND Corporation. Same report, recommendations. — “Focus on the problem, not the technology: Successful projects are laser-focused on the problem to be solved, not the technology used to solve it.” https://www.rand.org/pubs/research_reports/RRA2680-1.html
- KPMG Australia. “KPMG Australia shows disciplined performance in unpredictable FY25: releases annual impact report” (11 Aug 2025). — “The market environment for the Consulting business was impacted by a significant reduction in the government use of consultants, as well as the broader economic slowdown, with revenues down 18% for the year.” https://kpmg.com/au/en/media/media-releases/2025/08/kpmg-releases-annual-impact-report.html
- Nielsen Norman Group (Jakob Nielsen). “Progressive Disclosure.” — “Progressive disclosure defers advanced or rarely used features to a secondary screen, making applications easier to learn and less error-prone.” / “Initially, show users only a few of the most important options.” https://www.nngroup.com/articles/progressive-disclosure/
- DealHub. “What is a Sales POC (Proof of Concept)?” — “A POV typically involves the customer’s actual data and processes to highlight tangible benefits, while a POC is primarily focused on functionality, integration, and technical feasibility.” https://dealhub.io/glossary/sales-poc/
