Investors · Risks & controls
Risks & controls
Fixed fees at cost put delivery risk on us. These are the risks we see and what we do about each.
| Risk | What could go wrong | Controls |
|---|---|---|
| Delivery under fixed fees | Cost exceeds the fixed fee. Large IT projects have run 45% over budget on average.1 | 15% contingency in every fee. Scope fixed in the assessment contract. Change orders for anything outside it. Hours and inference measured per engagement from the first one. |
| Success-fee collection | The client disputes acceptance, or the fee arrives late. Our entire profit is this fee. | Acceptance criteria agreed up front in writing. Fee paid in tranches at defined milestones (such as rehearsal passed, go-live and hypercare exit), not as one payment at the end. The base is the client's current-path cost, set at contract. |
| Inference cost volatility | Prices change or usage runs above plan. Provider choice swings cost per engineer-month from $8K to $141K. | Provider-agnostic routing, metering by agent and engagement, spend caps, and model choice by task. Inference is budgeted at about $27K per engineer-month and replaced by metered results. |
| Agent accuracy | Agents state something wrong about a system or a number. | Findings cite evidence. Finance numbers are computed deterministically, not generated. Blinded evaluation sets. Human approval for any production change. |
| Data access and security | Sensitive data exposed, or an agent acts beyond its remit. | Four trust boundaries. Read-only pilot. Narrow, typed tools instead of open-ended access. Per-customer isolation. Minimal extraction through approved interfaces, and logged decisions. |
| SAP and integrator counter-moves | SAP's own assistants improve;2 integrators adopt AI and cut prices. | Assistants address tasks, not delivered programs, and no effort reduction is published yet. Integrators who adopt AI still bill hours, and our model prices below them at cost. SAP could be a channel partner. |
| Deadline dependence | The 2027 date moves or loses force; third-party support defers decisions.3 | Other triggers: unstable go-lives, older S/4HANA upgrades, M&A integrations, support exits. See Market. |
| Key people and talent concentration | The model depends on a few senior experts and the partners. | Phased hiring against contracted work. Cyril's functional network. Documented methods and reusable assets, so delivery does not live in individual heads. |
| IP and legal | Disputes over vendor software or reuse of work product. | We never reverse-engineer vendor software. We rebuild a client's rules from the client's own data, configuration, documents, observed process behavior and interviews. Contracts reserve our right to reuse non-client-specific assets while protecting client data and IP. |
| Board risk aversion on the freedom path | Boards will not move a ledger on a new model. | Staging: core first, in parallel, ledger last. Keep SAP as the system of record where regulation and audit require it. See Freedom. |
Notes
Early customers will want to see the controls working, so the design partner doubles as a security review. Financing terms are not committed. Diligence should cover customer discovery, a demo, the security architecture, team commitments, IP review and a monthly cash model.
Sources
- McKinsey & University of Oxford, 2012. Delivering large-scale IT projects.
- SAP News Center, 28 Oct 2026. Migration and Modernization Assistants.
- Rimini Street press release (vendor claim). Support through 2040.