Andrew RadosevichSignal 1: a written success metric before work begins
A real advisor will not start until the outcome is defined in writing. That means a baseline number, a target number, and the date by which the target is measured. Examples: lead routing time from 4 hours to under 5 minutes by day 30, claims submission throughput up 70% by day 60. If the proposal does not include a number, you are buying activity, not an outcome.
Signal 2: fixed-scope pricing tied to outcomes
Advisors price by package because the scope is the install, not the hours. A Diagnostic, a pilot, a program, an ongoing governance engagement, each with a fixed scope and price. Hourly billing on this kind of work tends to reward drift. If the pricing model is hours times rate, you are likely hiring a consultant on a retainer.
Signal 3: installation in production is the final deliverable
Read the statement of work. The final deliverable should be a workflow running in production, integrated with your existing systems, with monitoring and a documented runbook. If the final deliverable is a report, a roadmap, a deck, or a strategic recommendation, the install is not included.
Signal 4: a named internal owner trained at handover
Every advisor engagement should end with an internal person who can operate the system. That requires a training session, written runbooks, a monitoring dashboard, and a 30 to 90 day governance review cadence. Without an internal owner, you are dependent on the advisor forever, which is the opposite of the advisor model.
Signal 5: published case proof from businesses your size
Ask for two case studies from businesses within 50% of your revenue. Enterprise pilots do not translate to SMB constraints. If the only proof points are large companies, the operating model and pricing structure probably do not fit. SMB advisors should be able to name SMB clients and SMB outcomes.
Red flags to walk away from
Vague deliverables, no written success metric, hourly billing on implementation work, deliverables that end at a recommendation, no handover plan, no internal owner identified, no case proof at your revenue size, or a proposal that uses the word advisor but the SOW describes a report. Any two of these and you are not hiring an advisor.
Seven questions to ask on the first call
Ask these in order. One, show me a workflow you shipped for a business under $25M in revenue. Two, what is your fixed-fee Diagnostic price? Three, how do you measure ROI on an engagement? Four, which tools do you implement most often, and why? Five, what is the exit playbook, meaning when do we stop needing you? Six, how do you handle compliance and data privacy during the Diagnostic? Seven, can I speak with a reference at a company my size? A vague answer on any one of these is enough to keep looking.
What a good engagement structure looks like
Free readiness assessment, then a paid fixed-fee Diagnostic with a written 90-day plan that stands on its own, then an optional fixed-fee pilot that ships one production workflow in 30 to 60 days, then optional ongoing advisory. Each stage should be independently valuable and no stage should require the next. If the proposal only makes sense as a bundle, the structure was built for the firm rather than for you.
Common questions
How much should an AI advisor cost?
Fixed-fee entry points typically run $7,500 to $10,000 for a Diagnostic and $7,500 to $15,000 for a single workflow install. Programs scope after the Diagnostic so the price is tied to real workflows. If the first conversation ends with an hourly rate and no scope, the firm is not operating on an advisor model.
Should the advisor have industry experience in my vertical?
Helpful but not required. Workflow patterns repeat across industries: document processing, routing, scheduling, reporting, support triage. What matters more is whether the advisor has installed similar workflow types at similar company size. Industry depth becomes critical only in regulated environments like healthcare or legal.
How long does the engagement take?
A Diagnostic typically runs two weeks. A single workflow install runs three to four weeks. A multi-workflow program runs 60 to 180 days. If the timeline is open-ended, the scope is not defined, which means you are not on an advisor engagement yet.
What happens after the engagement ends?
You should have a live system, a trained internal owner, runbooks, a monitoring dashboard, and a written roadmap for the next two to three workflows in priority order. Optional ongoing governance covers monthly oversight, expansion, and vendor management for teams running multiple AI systems.
Should I work with a big-name consulting firm instead?
For most $1M to $25M businesses, no. Large firms are priced for enterprise budgets and staff engagements with junior consultants. Smaller founder-led firms that put senior operators on every engagement fit the constraints better.
What if a firm cannot show case proof at my size?
Walk. The distance between an enterprise AI pilot and a $5M business workflow is large, and enterprise techniques tend to be over-engineered for SMB constraints. You want proof from your weight class.
How much should the Diagnostic cost?
$5,000 to $10,000 fixed fee is the right range for a two to three week Diagnostic that produces a 90-day plan. Below that, the firm cannot afford to do real work. Well above it usually signals enterprise-shaped delivery, which is the wrong shape for an SMB.