peterbamuhigire SaaS Agent Moat & Wrapper Risk Skill
Overview
In 2025-2026, most "agent companies" are GPT wrappers in trench coats. The single most consequential question an agent-fund investor asks is: "If OpenAI / Anthropic / Google ships this capability directly next quarter, what's left of your business?" If the answer is "we'd be in trouble", the company is a wrapper and the valuation collapses to a fraction of the headline ARR multiple. If the answer is "we'd still win because of X, Y, Z" with evidence, the company is a defensible agent business and earns the agent premium.
This skill installs the moat-vs-wrapper discipline. Builds on saas-ai-moat-and-defensibility (which runs the 7-question AI moat test) and adds the agent-specific dimensions and the wrapper-risk-from-foundation-provider vector as its own first-class test.
Use When
- A SaaS plan claims an agent or multi-agent product is defensible
- Section 06 is being built for an agent-product plan
- A foundation-model provider has shipped a capability that overlaps the agent business
- An investor's technical diligence on agent defensibility is upcoming
- A founder is choosing between agent positioning ("we're an agent platform") and vertical positioning ("we're a vertical X agent")
- The plan asks for an agent valuation premium that implicitly assumes moat
- The plan must pass
meta-agent-bankability-and-investor-readiness
Do Not Use When
- The product is an AI feature inside SaaS without agentic action — use
saas-ai-moat-and-defensibility - The agent is internal-efficiency only with no customer-facing defensibility to test
- The plan is pre-PMF — moat claims are forward-looking, but discipline still applies; flag claims as "in development"
Required Inputs
- Agent architecture (planner / worker / critic; tool registry; data flows)
- Tool inventory — which tools are proprietary, which are third-party, which are commoditisable
- Action-data inventory — what action data accrues per customer (prompts, completions, tool invocations, outcomes, interventions, human-correction signals)
- Integration depth per customer — system-of-record integration, identity integration, audit-log integration, billing integration
- Eval-loop maturity — offline + online evals; production sampling; human-correction loops; eval-set proprietary or public
- Customer-trust evidence — references, regulatory clearance, audit acceptance, public-sector references
- Switching-cost evidence — data migration friction, integration switching, retraining cost, contractual lock-in
- Distribution channels — direct, partner, marketplace, embedded
- Foundation-model provider trajectory — is OpenAI / Anthropic / Google shipping in your category? GA / preview / rumoured?
- Competitor moat claims
Workflow
- List the moat claims in the plan — every place the agent is presented as competitive differentiation.
- Run the 8-question agent moat test per
references/saas-agent-moats-and-wrapper-risk-checklist.md:- Q1 — Proprietary tools — are the tools the agent uses proprietary (your APIs, your integrations, your data sources) or commoditisable?
- Q2 — Proprietary action data — does each customer interaction generate action data that improves the agent uniquely for you?
- Q3 — Integration depth — how deeply embedded is the agent in customer system-of-record, identity, audit, billing?
- Q4 — Eval-loop — do you have a real eval suite running in production with human-correction signals improving the agent over time?
- Q5 — Customer-trust / brand — are you the trusted agent vendor in this vertical / region?
- Q6 — Regulatory clearance — do you have explicit regulator engagement, audit acceptance, sectoral approval that competitors lack?
- Q7 — Switching-cost — data, integration, training that locks customer in (or, conversely, contractual lock-in alone — weak)?
- Q8 — Distribution — channel, partnership, embedded-in-larger-platform reach that competitors cannot match?
- Score 0-3 per dimension; max 24. Scoring guide in checklist.
- Apply the wrapper-risk detector per
references/agent-moats-vs-wrapper-risk.md:- Pure prompt cleverness — clever system prompt over GPT-4o/5 with no tools / data / integration / eval moat = wrapper
- Public-tool only — agent uses only public APIs and public data = wrapper
- Generic orchestration — wraps LangChain / CrewAI / AutoGen with no proprietary layer = wrapper
- Foundation-model partnership claim — "we partner with OpenAI / Anthropic" = wrapper (everyone has API access)
- Demo without deployment — agent runs in demos but no production customers = unproven; flag as risk
- Single-customer custom build — looks like a services business, not an agent product = services, not product
- No eval-loop — agent ships and never improves with customer data = wrapper
- Foundation-model platform risk — explicitly map: if OpenAI / Anthropic / Google / Meta / Amazon ship in our category next quarter, what survives? For each surviving piece, document why.
- Wardley-map the agent components — place planner / workers / critic / tools / eval-loop / observability / action-data on Genesis -> Custom -> Product -> Commodity. Components in Commodity cannot be your moat. Your moat must be in Custom / Product layers with explicit "stay defensible" logic.
- Compute the moat-vs-wrapper score — 0-24 from the rubric:
- 0-8: wrapper; valuation discount applies; recommend repositioning or repackaging
- 9-14: real but incomplete moat; valuation neutral; needs deliberate deepening
- 15-19: strong moat; agent premium territory
- 20-24: rare; strong defensibility across multiple dimensions
- Write the moat-or-wrapper thesis paragraph — one paragraph an experienced operator would not call marketing language. Must answer: what is proprietary, what accrues, what would survive a foundation-model commoditisation event, what is the customer-switching cost.
- Wire to valuation (
meta-agent-valuation-adjustments), to risk (saas-agent-risk-and-stress-test), and to executive summary (saas-agent-executive-summary-block.md).
Quality Bar
- Every agent-moat claim tested through the 8-question rubric
- Wrapper-risk detector applied; weak claims explicitly downgraded or retracted
- Foundation-model platform risk explicitly named with surviving-pieces inventory
- Wardley-map placement done for each agent component
- Moat-or-wrapper score computed; not gamed
- Thesis paragraph free of marketing language; evidence-based
- Cross-referenced to valuation, risk register, executive summary
- A sceptical agent-fund investor would not call the moat thesis "promotional"
Anti-Patterns
- "We're not a wrapper because we're vertical" — vertical alone is not a moat unless paired with proprietary tools / data / integration / eval / regulatory
- "We have a proprietary system prompt" — system prompts are not defensible (extractable, replicable)
- "We use a fine-tuned model" — one-time fine-tune without ongoing data accrual is not a moat
- "We partner with OpenAI / Anthropic" — API access is not a moat
- "LangChain / CrewAI / AutoGen is our framework" — wrapping commoditised orchestration is not a moat
- Mentioning eval without showing an eval suite + scores + improvement trajectory
- Skipping the foundation-model commoditisation question
- Wardley-mapping with all components in Custom (handwaving the actually-commoditised pieces)
- "Switching cost is high because customers signed annual contracts" — contractual lock-in alone is weak
Outputs
- Moat-claim inventory
- 8-question rubric score (0-24)
- Wrapper-risk detection list (claims rejected or downgraded)
- Foundation-model commoditisation analysis (what survives if provider ships)
- Wardley-map placement of agent components
- Moat-or-wrapper score and thesis paragraph
- Cross-reference to valuation premium / wrapper discount
Living-Plan Cadence Defaults
| Element | Cadence | Owner | Variance threshold |
|---|---|---|---|
| Foundation-model commoditisation watch | monthly | Head of AI / CTO | provider ships competing capability |
| Competitor agent claim scan | monthly | Head of Strategy | new entrant with similar claim |
| Action-data accrual evidence | monthly | Head of Product | accrual rate <plan |
| Eval-loop trajectory | monthly | Eval Engineer | eval improvement stalled 60 days |
| Tool-registry proprietary share | quarterly | Tool Engineer | proprietary share falling |
| Integration depth per customer | quarterly | Head of CS | shallow integrations on >30% customers |
| Moat-vs-wrapper score reassessment | quarterly | CEO + Head of Strategy | -3 points |
| Wardley-map refresh | annual | Head of Strategy | structural shift |
| Regulatory clearance maintenance | quarterly | Compliance + Legal | clearance lapse risk |
References
references/saas-agent-moats-and-wrapper-risk-checklist.md— 8-question rubric + wrapper-risk catalogue + Wardley placementreferences/agent-moats-vs-wrapper-risk.md— compact rubric (lives at06-competitive-analysis/references/)skills/06-competitive-analysis/saas-ai-moat-and-defensibility/SKILL.md— AI moat parentskills/06-competitive-analysis/SKILL.md— generic competitive analysisskills/meta-agent-valuation-adjustments/SKILL.md— consumes the moat scorebook-extractions/walling-saas-playbook-extraction.md— moat disciplinebook-extractions/agent-products-business-plan-audit-2026.md— agent audit
Africa / Uganda Application Notes
- Local-language agent moat is genuinely available in African markets — agents that work in Swahili, Hausa, Yoruba, Amharic, Igbo, Zulu, Luganda, Lingala, Wolof, Tigrinya at production quality are scarce and defensible. Pair with local TTS / STT (voice-IVR / WhatsApp-voice).
- Sovereign-AI / data-residency moat — agents that can demonstrably run on in-country compute with in-country data are procurement-favoured in KE, NG, ZA, RW, EG, UG public-sector and regulated sectors.
- Integration depth in African workflows — mobile-money rails (M-Pesa, MoMo, Airtel Money, Wave, Orange Money), USSD aggregators, WhatsApp Business API channel, identity systems (NIN / Huduma / Hudumba / Aadhaar-equivalents) are real integration moats; OpenAI / Anthropic / Google direct-ship cannot replicate the integration layer quickly.
- Regulatory clearance moat in African sectors — KE CMA fintech sandbox, KE ODPC, NG SEC / CBN / NITDA / NDPC, ZA FSCA / Information Regulator, RW BNR / NCSA, UG BoU / NITA-U / PDPO — agent vendors with explicit engagement and accepted-audit-log status have a real moat against foreign agents.
- Customer-trust moat — public-sector and regulated-sector procurement in Africa often disqualifies foreign agent vendors; local accountability matters disproportionately.
- Distribution moat in African verticals — 50 deployed SACCOs, 200 deployed clinics, 30 deployed schools, 5 deployed government agencies are real moats that foreign agents cannot replicate quickly.
- Wrapper risk in Africa — many "AI agent" startups in 2025-2026 are pure wrappers; the moat-vs-wrapper rubric must be applied strictly to avoid funding-stage embarrassment when DFI / IFC / AfDB technical diligence runs.
July 2026 Portable Contract
Required Inputs
| Input artefact | Source/provider | Required | Behaviour when absent |
|---|---|---|---|
| Named competitors and substitutes, customer decision criteria, observed offers, and defensibility evidence for saas agent moat and wrapper risk | Customer research, competitor primary sources, product data, and technical records | Yes | If absent, competitor or moat evidence is unavailable, state the untested claim and design a verification step rather than awarding a score. |
| Finalised business brief, target reader, country, and stage | Client intake and engagement owner | Yes | Stop section decisions and route the missing context to client intake. |
| Reconciled upstream assumptions that this section consumes | Named pipeline owners | Conditional | Record the dependency, affected claim, owner, and recovery step; do not substitute an invented value. |
Outputs
| Artefact | Consumer | Observable acceptance condition |
|---|---|---|
| Scored moat or wrapper-risk thesis with rejected claims | Plan author and target decision-maker | The artefact answers the section decision and traces each material conclusion to the supplied evidence. |
| saas agent moat and wrapper risk exception and handoff note | Downstream section owners | Every blocked or conditional item names its consequence, owner, evidence request, and restart condition. |
| saas agent moat and wrapper risk release record | Reviewer or plan assembler | Records the checks completed, failures, unassessed items, professional review required, and release state. |
Evidence Produced
| Evidence | Format | Acceptance condition |
|---|---|---|
| Rubric score, proprietary-asset proof, commoditisation test, and claim downgrade log | Source-linked table, calculation, or annotated prose | The evidence is reproducible from named inputs and distinguishes verified fact, management assumption, and inference. |
| saas agent moat and wrapper risk decision record | Decision note | States the selected action, rejected credible alternative, countercase, rationale, and risk accepted or avoided. |
| saas agent moat and wrapper risk review trace | Gate entry | Identifies the date, input versions, reviewer role, failed checks, recovery owner, and any check that remains not assessed. |
Capability and Permission Boundaries
For saas agent moat and wrapper risk, the controlling focus is proprietary tools, action data, integration depth, evaluation loops, switching costs, and provider commoditisation. This skill may inspect public and supplied competitive evidence in read-only mode; it may not misrepresent a competitor, scrape outside authorised terms, or assert proprietary advantage without proof. Its normal mode is read-only analysis and drafting. Any mutation, external communication, spending, certification, or professional conclusion outside that boundary requires explicit authority and must remain traceable to the approving role.
Degraded Mode
For saas agent moat and wrapper risk, loss of evidence about proprietary tools, action data, integration depth, evaluation loops, switching costs, and provider commoditisation activates degraded mode. If the controlling saas agent moat and wrapper risk evidence is unavailable, the same boundary applies. When competitor or moat evidence is unavailable, state the untested claim and design a verification step rather than awarding a score. Return the verified subset, label the affected decision qualified or not assessed, explain the downstream consequence, and state the smallest evidence request or authorised action that permits recovery. Do not convert the missing check into a pass.
Decision Rules
| Choice or condition | Action | Failure or risk avoided |
|---|---|---|
| For saas agent moat and wrapper risk, the claimed advantage is readily copied or depends only on a third-party model or framework | downgrade the claim, test what survives commoditisation, and reposition around evidenced assets or distribution | Promotional differentiation collapses under customer or investor diligence |
| For saas agent moat and wrapper risk, A current legal, regulatory, tax, accounting, market, or platform claim controls the saas agent moat and wrapper risk decision | Verify the controlling source, effective date, jurisdiction, and reviewer status before release | Stale external facts become permanent plan assumptions |
| For saas agent moat and wrapper risk, The evidence reconciles with neighbouring sections and the countercase does not overturn the choice | Complete scored moat or wrapper-risk thesis with rejected claims, attach the evidence and release record, and hand off named dependencies | Premature release and repeated downstream rework |
Workflow
- Define the exact saas agent moat and wrapper risk decision, intended reader, jurisdiction, business stage, and permission boundary.
- Collect named competitors and substitutes, customer decision criteria, observed offers, and defensibility evidence and map each material conclusion to its source; stop the affected conclusion when an input could change it.
- Apply the specialist methods and directly linked references already contained in this skill, retaining its domain thresholds, calculations, and Uganda or East Africa context where applicable.
- Compare the credible alternatives, test the countercase and failure path, and apply the decision table rather than selecting a template default.
- Produce scored moat or wrapper-risk thesis with rejected claims with the evidence, exception, and handoff records; reconcile every shared assumption with its owning section.
- Run the section quality checks, applicable finance or professional review, and anti-slop gate. If a gate fails, correct the evidence or decision and return to the responsible step.
Quality Standards
- Scored moat or wrapper-risk thesis with rejected claims must answer a real decision for the named bank, investor, DFI, grant, board, or strategic-partner reader.
- Rubric score, proprietary-asset proof, commoditisation test, and claim downgrade log must be source-linked, dated where facts can change, and sufficient for another reviewer to reproduce the conclusion.
- The section exposes its countercase, stop condition, recovery action, and effect on neighbouring sections.
- No unavailable source, calculation, tool, or professional review is reported as passed; finance and statutory judgements follow the governing doctrine.
- Language remains specific to saas agent moat and wrapper risk, uses British English naturally, and passes the repository anti-slop gate without promotional filler.
Anti-Patterns
- In saas agent moat and wrapper risk, treating an unavailable named competitors and substitutes, customer decision criteria, observed offers, and defensibility evidence as confirmed. Correction: qualify the affected conclusion and issue the named evidence request.
- Producing scored moat or wrapper-risk thesis with rejected claims that restates the brief but makes no choice. Correction: record the choice, rejected alternative, rationale, countercase, and implication.
- Ignoring a conflicting upstream assumption. Correction: return it to its owning section and resume only from a reconciled version.
- Reporting an unavailable check as passed. Correction: mark it not assessed and narrow the release state.
- Claiming compliance, assurance, bankability, or investor readiness from narrative quality. Correction: run the applicable gate and retain its evidence.
- Copying the worked example into a client plan. Correction: use the method only and replace every fact with verified engagement evidence.
Worked Example
An agent uses public APIs and a prompt but has no proprietary action data or deep integration. Score the exposed dimensions accordingly and describe it as wrapper risk until deployment evidence changes the result.
References
- Use the verified project evidence register and the owning upstream pipeline section for saas agent moat and wrapper risk; no local deep-dive reference is declared.
- For saas agent moat and wrapper risk claims involving money, tax, grants, reserves, revenue, cost, valuation, or financial statements, apply the Chwezi finance doctrine and record the required professional-review state; illustrative figures never become client facts.