What an AI Opportunity Assessment Includes - and What It Should Cost You

An AI opportunity assessment is the discovery stage of AI adoption packaged as a standalone, fixed-scope engagement: interviews with your team, a scored inventory of automation candidates, a data readiness check, and a prioritized roadmap with one-page briefs for the first pilots - typically delivered in a few weeks.

Its purpose is to replace opinion with evidence before money is committed to building anything. Done well, it is also the cheapest exit ramp in AI adoption: if the honest finding is that your workflows are not ready, or that a simple process fix beats an AI project, a good assessment says so.

Five artifacts define a complete assessment. The opportunity inventory: every candidate workflow found in interviews, each described with owner, frequency, effort, and systems involved. The scoring: each candidate rated on volume, rule-clarity, data readiness, verification cost, value, and risk - with the reasoning visible, not just the grades. The data readiness memo: where the needed information lives, what is accessible, what is ambiguous, and what personal data requires treatment under the LGPD. The roadmap: candidates sequenced into now, next, and later, with dependencies. And the pilot briefs: one page per recommended pilot stating the workflow, the AI's role, the human review point, the data required, and the measurable acceptance criterion.

Each artifact stands on its own. Whoever implements - the assessing consultancy, another vendor, or your own team - the documents are the specification.

It is not a tool selection exercise - naming products on day one is a sign the conclusion preceded the analysis. It is not a transformation program - a document promising to reinvent the company is a sales artifact, not an assessment. And it is not a technology audit of your IT estate - the unit of analysis is the workflow, not the infrastructure.

The boundary matters commercially: an assessment should be priced and scoped so you can walk away at the end with full value received. If the deliverables only make sense if you also buy the implementation from the same vendor, you are reading a proposal, not an assessment.

A typical shape for a small or mid-sized company: Week 1 - structured interviews across functions, focused on repetitive information work; the inventory takes form. Week 2 - scoring workshops with the people who do the work, plus the data readiness checks on the leading candidates: testing real cases against the recorded data and confirming system access. Week 3 - prioritization, pilot brief drafting, and a findings session where the reasoning is challenged openly before the report is finalized.

Your team's time investment is real but bounded: interviews and workshops, typically an hour or two per participant. The assessment consumes attention, not operations.

Four factors move the price of an assessment: the number of functions and workflows in scope, the number of systems whose data readiness must be verified, the depth of privacy analysis your data requires, and whether the engagement is delivered in one language or bilingually.

Structurally, assessments are fixed-price for a defined scope - that is the point, and it is what makes them comparable across providers. Two practical rules: a quote should state exactly which deliverables from the list above are included, and the assessment should be a small fraction of what the first implementation would cost - it is the de-risking step, not the project.

Three things, concretely. Decide: approve or reject specific pilots on evidence - each brief carries its value case, risk, and acceptance criterion. Commission: hand any competent implementer the briefs and the readiness memo as a specification, and hold the work to the written criteria. And repeat: the scoring method itself transfers - your team can rerun it on new candidates next quarter without external help.

If you cannot do those three things with the documents you received, the assessment was incomplete regardless of how polished the presentation was.

Watch for five. Tool names before workflow analysis. Interviews only with leadership - the repetitive work lives below the org chart's top layer, and so do the real candidates. No data readiness check - recommendations made without testing whether the data supports them are guesses. No acceptance criteria in the pilot briefs - "implement AI for customer service" is not a brief. And uniform enthusiasm - a scored inventory in which nothing is rated "not worth automating" means the scoring was decorative.

The common thread: a real assessment produces decisions you could defend to a skeptical board member, with the evidence attached.

The assessment pays for itself through two mechanisms, both measurable. Avoided waste: every seductive-but-unready project it demotes - wrong data, unverifiable output, thin value - saves multiples of the assessment's cost in implementation spend that would have disappointed. And compressed time-to-value: the first pilot starts with a specification instead of a debate, which typically removes weeks of alignment churn.

Demand this framing from any provider: the assessment's product is decision quality. Its success is visible in the first pilot's measured result against the written criterion - not in the thickness of the report.

ConsultatechAI delivers assessments in the exact shape this article describes - inventory, visible scoring, data readiness memo with LGPD treatment, roadmap, and pilot briefs with acceptance criteria - bilingually in Portuguese and English, for companies in Brazil and internationally.

We hold ourselves to the same warning signs: no tool names before analysis, interviews below the leadership layer, and at least part of every honest inventory marked "do not automate yet." The assessment is priced as a standalone deliverable you own outright - whether or not we build what it recommends.