CVAi is a guided workspace where a team of AI agents runs a CVA consulting methodology end to end: they read your approved sources, draft every analysis, and build the review-ready deck and model — while your experts validate each material step at explicit gates. Not a chatbot. A method that runs.
The most valuable consulting work is methodology-driven and recurring: market diagnostics, country action plans, business cases, due diligences — the same proven method, replayed across countries, business units and deals. And every replay pays the same three taxes.
The frames exist, the benchmarks exist, the storyline exists. Yet each new country or BU re-derives them by hand: weeks of senior time spent on work the method already knows how to do.
The method is only as good as the people applying it that week. Two teams, same methodology, different rigor: the client sees the difference, and so do you.
Sources, assumptions, discarded options, the why behind every number: none of it survives delivery. The next team starts cold, and nothing is auditable six months later.
Your team opens a mission, uploads the corpus, and is guided through the method stage by stage. Agents extract the evidence, run the diagnostics, score the gaps, shape the levers and draft the deliverables. Nothing needs prompting; nothing starts from a blank page. Your experts do exactly one thing: decide — approve sources, validate evidence, sign gates.
Free-form chat invents. In CVAi, no agent can write a material claim without an approved source or an explicitly labeled assumption — the workspace blocks it mechanically, and honest states are shown as such: “Needs validation”, “Not enough evidence”, never a confident guess.
Framing, benchmarks, diagnostics, scoring, workshops, plan: your methodology's stages, gates and registers, encoded once and enforced on every mission
Every source is parsed for real and individually approved. Every claim anchors to an exact excerpt: the yellow highlight is three clicks from any number
Board-grade decks and formula-live models rendered deterministically in your client's template, plus interim steerco and workshop packs between gates
Explicit gates at every methodology moment. Approvals carry initials and land in an append-only journal. Uploaded is not approved; drafted is not final
Three roles, cleanly separated — with a hard gate in the middle. That gate is not a disclaimer, it is the product's core mechanism: your organisation's judgment enters the work in a form that is explicit, structured and journalled.
An orchestrator routes the stages. Specialist agents qualify sources, extract quote-anchored evidence, flag contradictions between sources, run the analyses and draft every deliverable. They never guess: gaps are flagged, not filled.
AI executesScope, source use, evidence, each analysis stage, the final review: nothing advances until a named reviewer signs it with initials. Your team's judgment enters the model here — traced, timestamped, reversible by reopening the gate.
You decideDecks and models are built by code, not by a model: schema-checked, formula-aware, brand-exact. Mechanical checks block any content citing unapproved material, and the final gate certifies the actual rendered files.
Code certifiesMost AI work fails at the partner's desk because nobody can say where a number came from. CVAi is engineered in the opposite direction: the audit trail is the product. Evidence first, prose second, and a journal that answers “who approved this, based on what” months later.
This is the exact control your reviewers use before anything leaves the workspace. The review-ready gate certifies the rendered files — not intentions, not drafts. Until it is signed, downloads stay locked. Sign it below.
Everything in CVAi that matters works exactly like this: sources, evidence, assumptions, contradictions, every stage, the final release. Machines prepare and check; only a named human moves the state. And reopening is one click — approval is a decision, not a ratchet.
The end product is not a chat transcript. It is the same set of files your teams produce today — built deterministically, in your client's exact template, with every figure traceable.
The full storyline in fixed, methodology-defined sections plus appendix depth: real charts (heatmaps, timelines, waterfalls, donuts), your client's template, fonts and logo grammar, and mechanical slide-fit rules so nothing ever overflows or shrinks to fine print.
A workbook with live, auditable formulas: inputs, scoring, business case, sensitivities, prioritization, action plan — plus the audit sheets: source register, evidence references, contradictions and the full validation log.
A gated interim pack for steering committees: where the mission stands, what is validated, what is blocked and why — honest status, never a pretty façade over unfinished work.
Pre-reads and seed content for ideation and prioritization sessions with the client team: the evidence base, the gaps, first hypotheses to react to — so workshops start at altitude, not at zero.
Slides and sheets are built by deterministic renderers from validated content — the same input always produces the same file. Agents author the thinking; code does the typesetting. That is why the formatting never drifts and the formulas always add up.
The first CVAi tool runs a full vaccine-coverage strategy methodology in production today, end to end on real client sources. But the engine underneath carries zero industry assumptions — it only asks for ordered stages, human gates and evidence-linked registers. Your methodology defines the rest.
The commercial proposal that describes your methodology, plus past project deliverables: final decks, models, steerco packs. The system reads them and reverse-engineers the method — stages, gates, frameworks, deliverable formats, client style.
AI ingestsYou receive the methodology on a single parameter sheet, each answer citing the slide or tab it came from, plus the open questions. You correct and sign. This is the build's one human gate — and its contract.
You decideThe engine is forked, the domain layer regenerated from your sheet, the client brand applied, the full verification ladder passed, and the tool deployed — in days, inheriting every hardening the engine has ever received.
Engine buildsEach tool runs as a dedicated instance with its own store and access gate. Your corpus stays in your perimeter; EU hosting or deployment inside your own infrastructure
Sources and deliverables are used for your missions and nothing else: never pooled across clients, never used to train models, never leaving the instance
Decisions and audit events are journalled in a ledger that can be extended but never silently edited. Who approved what, when and on which evidence is always answerable
Any claim in any deliverable resolves to its evidence card and the exact highlighted excerpt in the source document. Three clicks, no dead ends
Access-code gate on every instance; every validation carries its reviewer's initials. Machines propose — identified humans decide, and the journal remembers
Runs on Claude today; the agent layer is provider-portable and can be plugged into your own approved AI infrastructure by our partners
When something is blocked, the workspace says so and why — a missing source, an unapproved card, an unsigned gate. It never fakes progress: no simulated results, no invented citations, no silently filled gaps. That is a design law, enforced in code.
CVAi is deployed the way a decisive strategy engagement is run: senior-led, evidence-based, built around your teams. The deliverable is not a licence — it is your methodology, running as a capability your organisation owns, with CVA at your side when the stakes peak.
Anyone can wire an LLM to a slide generator. The value is in the method — what to ask, what evidence counts, when a human must decide, what a partner will sign. That is consulting judgment, and it is what CVA has been building for 35 years. CVAi is how it compounds instead of evaporating.
“An AI consultant is worth exactly the method inside it. We spent 35 years building the method. Now it runs, with your people holding the pen that signs.”Corporate Value Associates