Not Everything Called AI Is AI: An AI-Washing Audit of a 22-Item Product Roadmap
A B2B FinTech platform had 22 roadmap items marketed as "AI-powered" ahead of a funding round. Envion classified each one against a four-tier definition. Nine were deterministic business rules, six were reporting and aggregation, five were genuine machine learning or LLM features, and two were not feasible as described. Honest reclassification shortened the near-term roadmap, removed two items from an EU AI Act risk tier they did not belong in, and made the five real AI items properly resourced for the first time.
The challenge
The client's concern was commercial: competitors were shipping "AI" features and the roadmap needed to match. The risk, which surfaced in the first workshop, was that engineering had begun estimating several items as machine-learning work when they were in fact conditional logic — and were being resourced, scheduled and de-risked as if they required data science.
Two costs compound here. Internally, a rules engine estimated as an ML project attracts an ML timeline, an ML team and ML uncertainty for work a senior backend engineer could finish in a sprint. Externally, calling deterministic software "AI" in a regulated market invites scrutiny — from customers' procurement teams, from auditors, and increasingly from regulators.
Decision path
We applied a four-tier definition and forced every roadmap item into exactly one tier — no item was allowed to sit in two.
Tier 1 — deterministic logic and arithmetic: same input, always the same output, and a human can read the rule. Thresholds, scoring formulas, eligibility checks, validation, routing rules, weighted calculations. This is not AI; it is software, and it is often the correct answer. 9 items.
Tier 2 — analytics and reporting: aggregation, segmentation, dashboards, trend lines, cohort analysis. Valuable, well understood, and not AI. 6 items.
Tier 3 — statistical and machine learning: the system learns parameters from data rather than having them specified — anomaly detection on transaction patterns, propensity scoring, forecasting with learned seasonality. 3 items.
Tier 4 — generative and language models: LLM-based summarisation, extraction from unstructured documents, natural-language interfaces. 2 items.
Not feasible as scoped: two items required data the platform does not hold and could not lawfully collect. 2 items.
Envion contribution
Envion delivered the classification matrix, a revised roadmap, a one-page terminology standard, a regulatory exposure map and an investor-facing narrative — over three weeks with one AI strategy lead, one solution architect and one AI governance advisor.
Delivery: the nine Tier 1 items moved out of the data-science backlog into normal product engineering — seven of them shipped the following quarter (client-reported) because they had never been hard, only mislabelled. Resourcing: the five genuine AI items had collectively been sharing attention with seventeen others; concentrating the same budget on five gave two of them dedicated evaluation infrastructure — the difference between an AI feature and an AI demo.
Delivery
Regulatory exposure: under the EU AI Act, obligations attach to systems meeting the definition of an AI system and to their risk classification. Two items had been provisionally flagged as high-risk on the strength of their internal name alone; once documented as deterministic rules with published logic, they fell outside that framing — while one Tier 3 item that nobody had flagged turned out to warrant genuine attention.
Language: the one-page terminology standard defines what the company may call AI in customer-facing material, what it must call automation, and who signs off. It is now part of the marketing review.
Outcome and evidence
22 audited roadmap items became 5 genuine AI, 15 correctly renamed as logic or reporting, and 2 removed as not feasible. Seven of the nine Tier 1 items shipped the next quarter (client-reported), and two items incorrectly flagged as high-risk under the EU AI Act were corrected (client-validated).
Practical rules from the project: apply the explanation test — if a competent engineer can write the complete rule on a whiteboard, it is logic; ask where the parameters come from — specified by a human means not AI, learned from data means AI; beware the word "smart" — it is usually a threshold; do not treat reclassification as a demotion — deterministic logic is faster, cheaper, testable and often more accurate; write the terminology down before marketing does; and remember regulators read the behaviour, not the label — in both directions.
Evidence gate. This page publishes only what Envion's project records and client disclosure permissions support. Outcomes are added once verified against a baseline, a measurement period, and an approved source.
FAQ
Questions about this case
Facing a similar challenge?
If your roadmap is full of "AI-powered" items heading into a funding round or a compliance review, discuss a classification audit with Envion — three weeks to a roadmap you can defend line by line.
Discuss a Similar ChallengeKeep exploring
Similar case studies
Executive Technology Leadership
Support for high-stakes product and AI decisions
Bring senior technology leadership into the business when the roadmap is unclear, delivery is at risk, an AI initiative needs stronger ownership, or the company needs an experienced technical voice before hiring a permanent CTO.
Discuss Interim CTO SupportCore responsibilities
- Align product and technology priorities with business goals and measurable outcomes.
- Review architecture, delivery risks, data foundations, security needs, and AI readiness.
- Lead internal teams and external partners through a practical execution plan.
- Clarify team structure, ownership, decision rights, and delivery cadence.
- Support investor, board, partner, and due-diligence conversations with credible technical judgment.
New experience
Prompt-to-Page — try it right here
Describe the landing page you want, in your own words. We turn it into a finished page and email you a private link in 5–10 minutes — no briefs, no calls, $0 to see the result.
- Describe what you want to create.
- We structure, write, and compose the page.
- You receive a private link when it is ready.
Start with a sentence — the interactive builder takes it from there.
Generate My PageSafe, respectful content only. No obligation.