The AI Consultant Starter System.

Act as a workflow opportunity analyst. Interview me about repeated work, delays, re-entry, errors, and handoffs in my business. Ask one question at a time. Produce a shortlist ranked by frequency, time cost, business impact, reversibility, and data sensitivity. Do not recommend a tool until the workflow is clear.

Map my current software stack. For each tool, capture its purpose, owner, data entered, data produced, integrations, monthly cost, and failure points. Identify duplicate capabilities and manual transfers. Separate confirmed facts from assumptions and ask before filling any gap.

Assess one workflow for AI readiness. Check task stability, input quality, exception rate, privacy, permissions, consequences of error, current baseline, human ownership, and rollback. End with Ready, Needs preparation, or Do not automate yet, and explain the evidence for the rating.

Create an as-is process map from my description. List the trigger, each step, system used, owner, input, output, waiting time, decision, exception, and final result. Mark unknowns clearly. Then propose a smaller to-be pilot that changes only the minimum necessary steps.

Review this proposed AI workflow and design human checkpoints. For every consequential action, specify who reviews it, what evidence they see, approve/edit/reject options, timeout behavior, escalation path, and audit record. Prefer reversible drafts over automatic final actions.

Perform a least-privilege review for this workflow. List each data field and system permission requested, why it is needed, whether it can be removed or masked, retention needs, prohibited data, and who may access the result. Flag any credential sharing or unrestricted write access.

Write a plain-language automation specification for this workflow. Include trigger, required inputs, validation, deterministic rules, AI steps, integrations, outputs, exception queue, retries, ownership, logging, rollback, and acceptance criteria. Keep assumptions in a separate list for approval.

Create an operating prompt and a short SOP for this task. The prompt must define role, allowed sources, required output, uncertainty handling, prohibited actions, and escalation. The SOP must explain how a person starts, reviews, corrects, records, and stops the process.

Design UI test coverage for this workflow. Include the happy path, validation, empty state, duplicate submission, permissions, timeout, partial failure, recovery, accessibility basics, and mobile viewport. For each case provide preconditions, steps, expected result, evidence to capture, and cleanup.

Create a measurement baseline for this workflow without inventing numbers. Define volume, handling time, waiting time, error/rework rate, completion rate, operating cost, and customer impact. Provide a simple collection plan and the minimum sample needed before comparing a pilot.

Analyze this de-identified exception log. Group failures by cause, frequency, business impact, detectability, and recovery effort. Show representative examples, distinguish system defects from bad inputs and policy cases, and recommend what to fix, route to a human, or leave manual.

Plan a 30-day pilot for this workflow. Define scope, owner, users, sample data, safeguards, baseline, weekly milestones, acceptance thresholds, stop conditions, incident response, review cadence, and the final keep/improve/stop decision. Do not assume production access is approved.