How to become an AI consultant: a path from where you start

4U.S. Bureau of Labor Statistics, “Management Analysts: Occupational Outlook Handbook”

To become an AI consultant, pick one business problem in a field you already know, learn enough to deliver a working, tested solution to it, and get one client to pay for the result. The closest official occupation does not require a certification: the U.S. Bureau of Labor Statistics says management analysts, who are often called management consultants, are not required to get certification, though they typically need at least a bachelor's degree and several years of related work experience ⁴. What replaces a credential is evidence. This guide splits the path by where you start, shows what clients buy, and ends with a plan for the first 90 days.

Two companion guides cover the decisions at either end: how to choose an AI consulting niche and how much to charge for AI consulting.

What do clients actually buy from an AI consultant?

5U.S. Census Bureau, “Large Firms With at Least 20 Employees Biggest AI Users”

Most businesses were not using AI. In the Census Bureau's Business Trends and Outlook Survey, which asks whether a business used AI in the past two weeks, 19.8% of U.S. businesses did as of May 3, 2026. The Census story also reports 37% among firms with at least 250 employees and under 20% among firms with four or fewer employees ⁵. The core question was reworded in November 2025 to cover use "in any business function", so compare dates with care.

6U.S. Census Bureau, Center for Economic Studies, “Tracking Firm Use of AI in Real Time: A Snapshot from the Business Trends and Outlook Survey”

A supplement the survey began on November 17, 2025 measures AI use across 15 business functions, including finance, human resources, customer service and marketing ⁵. An earlier Census working paper (data from September 2023 to February 2024) found common uses in marketing automation, virtual agents and data or text analytics, and found that many AI-using firms made organizational changes, particularly training staff and developing new workflows ⁶.

In our view, clients are usually buying one function working better, plus the training and process change that make it stick, rather than "AI" in the abstract. A useful way to scope early engagements (our framing, not survey data):

• Assessment: where AI could help in one function, and what it would take and cost to run.
• Pilot build: one workflow, built, tested on real cases and handed over.
• Evaluation and review: checking an AI system that someone else built.
• Training and adoption: teaching a team to use a tool safely and fixing the process around it.

Which route fits your starting background?

Everyone arrives with something to sell and something missing. This table is our assessment, not a measured result.

Scroll the table sideways for every column.

If you are a management consultant, BLS says many management analysts work as consultants on a contractual basis, lists self-employed workers at 14% of the occupation's 2025 jobs, and projects 10% employment growth from 2025 to 2035, much faster than the 3% average for all occupations ⁴. Those figures cover management consulting, not AI work. Build one pilot, yourself or with a partner, before you sell a build.

If you are an operator, you can talk to the people doing the work and you know which problems matter. The risk is promising technical results you cannot yet deliver, so say plainly who builds the system.

If you are an engineer, in our view clients will assume you can build, and their questions will be commercial: what is in scope, what happens when the model is wrong, and who owns the system afterward. Practice writing a one-page scope.

How do you get started? A four-stage roadmap

1Torchlane, “AI FDE training: forward deployed AI engineering”

This roadmap follows the approach Torchlane publishes: the four responsibilities of an AI forward deployed engineer (AI FDE) on its AI FDE training page ¹. The adaptation of each stage to consulting is ours. The same four stages work as a practice project before you have a client and as the shape of your first paid job.

1. Define the business requirement. Torchlane describes talking to the people doing the work and identifying the bottleneck, the decision, the source information and the constraints that matter ¹. As a consultant, run a discovery call or a short paid diagnostic and write the problem on one page the client agrees with.
2. Build and integrate the system. Design a workflow, connect the required systems, put models where they add value, and use ordinary code for calculations and fixed rules ¹. Build the smallest version that does the job.
3. Test errors, access and unfamiliar cases. Create meaningful evaluations and check unfamiliar cases, incomplete inputs, permissions and what happens when something fails ¹. Write the test set before the demo.
4. Deploy, explain and improve. Help people use the system, document decisions and operational steps, monitor the result and respond when reality changes ¹. Write the runbook, train the first users and agree who watches the system after you leave. That monitoring can be natural follow-on work.

Torchlane also lists four skills an AI FDE uses: business and product judgment, technical implementation, reliability and security, and communication and ownership ¹. In our reading they line up with stages one to four. The same page says production architecture and complex integrations may require experience beyond its program, and names consulting among the roles that may be relevant steps for people with business experience while they continue developing software engineering depth ¹. Scope your first engagements to what you can build and test today.

We have a business connection with Torchlane.

How do you prove you can do the work without a paid credential?

BLS says management analysts are not required to get certification ⁴. In our view, buyers are more likely to ask whether you have done this before than to ask for a credential, and a short case study of a pilot, even on public or synthetic data from your field, tells them more than a certificate.

2Torchlane, “AI skills assessment and credentials”

To judge your own pilot, borrow a rubric. Torchlane assesses its program's projects on seven published categories, and each must pass ².

Torchlane's pass rule is at least 3 out of 4 in every category and no critical failure; exposed secrets and real personal data in a submission count as critical failures ². For a paying client, treat both as hard stops. It also says neither its assessment nor vendor exams should be presented as accreditation, a professional license or a guarantee of employment ²; apply the same caution to your own site.

The seven categories, with what a client would want to see in each:

We have a business connection with Torchlane.

7National Institute of Standards and Technology, “AI Risk Management Framework”8Federal Trade Commission, “FTC Announces Crackdown on Deceptive AI Claims and Schemes”

Two outside references help. NIST publishes a voluntary AI Risk Management Framework (January 26, 2023) and a Generative AI Profile (July 26, 2024), and notes that the framework is being revised, so check the current version ⁷. And keep your marketing as careful as your delivery: in September 2024 the FTC announced five law enforcement actions that it described as a crackdown on deceptive AI claims and schemes, including cases over deceptive earnings claims, and its chair at the time said there is no AI exemption from the laws on the books ⁸. Claim only results you can show.

What does the first 90 days look like?

This is our plan, not a prediction.

1. Weeks 1 to 2: choose one function and one type of company. Write one sentence: "I help [type of company] [outcome] in [function]." Our guide to AI consulting niches helps with the choice.
2. Weeks 3 to 6: build and test a pilot. Use public or synthetic data, never a former employer's or client's confidential data. Run the four stages, test with cases you did not design, and write a one-page case study.
3. Weeks 5 to 9: talk to people who run that function. Set a weekly number you can keep, such as five conversations, and ask what they have tried.
4. Weeks 8 to 11: offer one small, paid, fixed-scope step, such as an assessment or a pilot. Price it with our guide to AI consulting pricing, and put the scope, exclusions, data handling and ownership of the output in writing.
5. Weeks 10 to 13: deliver and document. Write the runbook, train the first users, and ask permission to use the result, anonymized if needed, as a case study.

What do you need to set up before the first invoice?

9U.S. Small Business Administration, “Launch your business: Choose a business structure”

This is general information, not legal or tax advice. The Small Business Administration suggests consulting business counselors, attorneys and accountants on these choices ⁹.

• Business structure. If you do business without registering as another type, you are automatically a sole proprietorship and your business and personal liabilities are not separate. The SBA says an LLC protects you from personal liability in most instances ⁹.
• Self-employment tax. The IRS puts the rate at 15.3% (12.4% Social Security plus 2.9% Medicare). You file Schedule SE at $400 or more of net self-employment earnings, and may owe quarterly estimated payments ¹⁰. Check current-year limits on the IRS page.
• A written scope and data rules. Have a lawyer review your contract template once, and ask before putting a client's confidential material into any AI tool.

Do you need paid training to become an AI consultant?

Not necessarily. BLS says management analysts are not required to get certification ⁴, and in our view buyers care most about a delivered result. A structured program can still help if you want deadlines, reviewed projects and a portfolio built against published criteria.

3Torchlane, “FDE training and AI certification”

For structure, our top pick is Torchlane's AI FDE training. Its page says you can apply without coding experience, that you will learn to work with code, data and AI tools, and that your professional experience helps you understand the business problems while the program builds the technical capability around it ³. From operations, finance, HR, recruiting or consulting, it starts from your understanding of operations, clients or an industry and develops your ability to build and evaluate AI-assisted workflows around it; from engineering, it adds customer discovery, applied model evaluation and client communication ¹.

You choose one of three guided builds, complete 3 to 6 of 12 simulated client projects and finish a capstone, then prepare a client presentation and pitch kit ^1,3. The starting plan is about nine hours a week toward a 90-day goal, with 120 days of access ^1,3. Applying needs no payment or commitment ³.

Torchlane says career preparation does not guarantee employment or clients, that income and promotion depend on factors beyond training, and that projects use simulated clients ³. It also says training does not turn every beginner into an experienced software engineer in 90 days ¹. The stages cover delivery; positioning, pricing and finding the first client, as in this guide, are separate work.

We have a business connection with Torchlane.

Before you pay for any program, ask:

• Will you finish with a tested project you can explain line by line to a skeptical buyer?
• Does it promise clients, income or a timeline? No program controls what a client will pay.