AI/4C Advisor turns an executive interview into a shortlist of AI project opportunities, each rated on the same five dimensions and grounded in a one-line rationale. We do not assert ROI, savings, or payback numbers for a project that does not exist yet — a rating a competent team could not defend on the day it is made is not worth pretending to compute. Every band below is what the assessment pass is actually held to, published so the rating means the same thing across every engagement.
Value toward the goals THIS executive stated in the profile — not abstract industry value. An opportunity that would transform a workflow the company does not care about is low impact here.
Directly advances a stated top goal or relieves a stated top pain.
Clearly helps a stated goal, but indirectly or within a limited scope.
Real but marginal value; does not map to anything the executive stated.
Effort and complexity for a competent small team to reach a working pilot at THIS company, given the systems and constraints in the profile. Not the difficulty of a full production rollout.
A well-trodden build with the pieces already in place.
Achievable but with real integration or data preparation work.
A serious build: deep integration, heavy data work, or organizational change.
How long until a first working version is in real users' hands under supervision — a pilot, not a production rollout. Assumes a competent small team starting from the company's current state per the profile.
A working pilot inside roughly two to six weeks.
Roughly a quarter: real setup work before the pilot can run.
Multiple quarters: foundations must be built before the idea is testable.
Whether the data this opportunity needs exists, is accessible, and is usable — derived from the profile's systems and workflows sections, not assumed.
The needed data exists, is accessible, and is usable roughly as-is.
The data exists but needs collection, cleanup, or access work first.
The data does not exist yet or is blocked — collection or policy work comes first.
How well the profile grounds this specific rating. Confidence qualifies the rating; it is not a rating of the opportunity itself.
Grounded in specific facts the executive stated or confirmed.
A reasonable inference from the profile, but key specifics are missing.
The profile is thin here — treat the rating as provisional.
This catalog exists so the generator sweeps every business function instead of missing obvious wins — the recommendations Advisor makes for your company are free to be bespoke ideas well beyond this list.
Incoming support requests are classified, routed to the right queue, and answered with a drafted reply a human approves. The highest-volume, best-understood first AI project in most companies.
Employees ask questions in plain language and get answers grounded in company policies, SOPs, and internal documents, with the source shown. Replaces "ask the person who knows" for routine questions.
Documents that arrive as PDFs, scans, or email attachments — orders, forms, statements — are read automatically and their fields land as structured data in the target system, with low-confidence items queued for a person.
Supplier invoices are extracted, matched against purchase orders or contracts, and routed for approval; mismatches and unknowns go to a human exception queue instead of a shared inbox.
New claims, applications, or cases are captured from submitted documents and checked for completeness on arrival, so applicants hear about missing items in minutes instead of after days in a queue.
Incoming contracts are compared against your standard positions: key clauses extracted, deviations flagged, and a summary drafted for counsel. A speed-up for legal review, never a replacement for it.
Meetings and calls are summarized into decisions, owners, and action items that land in the tracker or CRM — instead of evaporating or living in one person's notebook.
Routine outbound writing — status updates, renewal notices, vendor replies, follow-ups — is drafted from templates and context, with a person reviewing and sending. The person stays the sender; the blank page goes away.
Sales calls are summarized into CRM updates, next steps, and objection themes automatically — so the CRM reflects reality without reps spending evenings on data entry, and pipeline reviews argue from the same facts.
RFP questions are matched against a library of past responses and product facts, and first-draft answers are assembled for the bid team to edit — cutting days off response cycles for document-heavy sales.
Weekly and monthly rollups — ops reviews, board updates, client status reports — are assembled from source systems into a first draft, with the numbers pulled (never invented) and the narrative drafted for an owner to edit.
The mechanical parts of forecasting — chasing inputs, consolidating spreadsheets, drafting variance commentary against actuals — are automated, while the numbers and the final judgment stay with finance.
Every support reply, call, or case note is scored against your quality rubric instead of the 2% a QA team can sample — surfacing coaching opportunities and systemic issues while humans handle the consequences.
Outbound content or transactions — marketing claims, advisor communications, expense reports — are screened against written policy before they go out, flagging likely violations for human review.
Duplicate and inconsistent records across systems — customers spelled three ways, products under different codes — are matched and merged with confidence scores, with ambiguous cases queued for a person.
Competitor moves, market news, and regulatory changes in your space are monitored continuously and delivered as a short periodic digest with sources — replacing ad-hoc Googling with a consistent watchlist.
Job descriptions are drafted from role requirements, and applications are summarized into structured screening notes against the stated criteria — with humans making every advance/reject decision.
New hires and cross-trainees learn through an assistant grounded in your actual SOPs and policies — answering questions, walking through procedures, and quizzing understanding — instead of shadowing whoever is free.
Lifecycle messages — onboarding sequences, renewal notices, win-back campaigns — are tailored to each customer's segment, history, and behavior instead of one blast for everyone.
Your customers get an assistant inside your product — answering usage questions, guiding setup, or operating features conversationally. A product-differentiation play rather than an internal-efficiency one.
A repetitive multi-step process — vendor onboarding, order exception handling, account provisioning — is executed end-to-end by an agent that works across systems, pausing at defined human approval points.
Operational alerts — system incidents, quality deviations, fraud flags — are triaged automatically: clustered, matched against past incidents, and annotated with probable cause and the relevant runbook before a person picks them up.
Rubric rubric-v1 (2026-07-10) · Archetype catalog catalog-v1 (2026-07-10)
Advisor output is idea generation and planning support — not financial, legal, or procurement advice — and comes with no guarantees.