AI agent consulting for service firms: agents that do the work, approved by your people, run by us.

An AI agent consultant turns one recurring job into an agent your team can trust. They choose the job, write down what the agent may read and prepare, set the point where a person approves, build it in your tools and run it on real cases. The deliverable is a working agent, not a slide deck.

At the end of a sound engagement, you can point at five things:

1. The job card. One page: what the agent reads, what it prepares, who approves, what it hands back.
2. The agent itself, running in your email, calendar, CRM or shared drive.
3. The approval queue, where your people approve, edit or reject each output.
4. The written rules, built from the corrections your team validated.
5. The numbers, reviewed with you every month.

A consultant who cannot show you those five sold you advice about agents, not agents. If your question is wider, such as where AI fits in the firm as a whole, start with what an AI consulting firm does.

Owner-led service firms of 5 to 50 people where a partner still rebuilds the same text-heavy work every week: consulting practices, executive search, M&A advisory, interim management. The test has three parts. The work recurs, it follows your methods, and it waits for a partner's attention. Miss one, and an agent is the wrong tool.

The jobs that pass the test are rarely glamorous: a prospect list ranked for a partner to approve, a proposal drafted from your past proposals, a meeting brief built from your notes and CRM, the follow-up prepared after each call.

The honest answer is sometimes no. If someone in your team can configure, test and maintain an agent, buy the tool and save the fees. If the work happens twice a year, keep it manual. The Leverage Assessment says so in writing when it is the case.

Prepare, never commit. An agent may read the sources you list, draft, rank, summarise and remind. Anything that commits the firm, an email to a client, a proposal, a price, a post, waits for the yes of a person you name. Write that line first, before you choose any tool or any consultant.

We write it as a job card, one per agent, in four lines: reads (the sources it may use, and nothing else), prepares (the exact output, in your template), approves (the person who says yes, and what they check) and hands back (the cases it returns to a person instead of guessing).

Example, a scenario. A consulting firm of fifteen people wants help with proposals. Its job card reads:

• Reads: the firm's past proposals in the shared drive and the notes from the discovery call.
• Prepares: a first draft in the firm's proposal template, with the sources it used.
• Approves: the partner who owns the client relationship, before anything is sent.
• Hands back: a fee outside the firm's usual range, a service it has never sold, a missing call note.

The trap. The fourth line is the one most builds skip. An agent without it fills the gap with a guess, and the partner hears about it from the client. Copy the four lines for one of your own processes and paste them into your brief: it is the fastest route to a precise first call.

The whole job, in order. A Leverage Assessment decides which agent is worth building, a bounded test runs on your past cases, the build goes into your tools, and then we operate the agents every working day. Your team approves every external send and keeps the final word at each step: continue, adjust or stop.

Operating is where agent work is won or lost, so here is ours:

• We watch the runs, fix what breaks and train the people who work with the agents.
• Each correction your team validates becomes a written rule, so the next draft needs fewer of them.
• A monthly working session on the numbers with a dedicated consultant, and a new agent installed every quarter.
• Your data stays in an installation dedicated to your firm, hosted in Germany at Hetzner. None of it trains a model, and you keep a full export if you stop.

One published example: at Asia-Connect, an executive search firm, AI Jungle installed and runs a business-development agent, and a partner approves every message before it goes out. The five steps are on the method page.

The first one is free. You answer three questions about one process, with no account, and Loïc Guyon, who founded AI Jungle, reads your brief before a 30-minute call. The call answers one question: does that process deserve an agent, a simpler fix or nothing at all?

The paid step is the Leverage Assessment: one consultant, two 45-minute sessions and a report reviewed by a person in between. It ends with a recommendation, the alternatives and an action plan, and it never commits you to a build.

Ask for proof you can check, not a demo. A credible company writes the job card before it quotes, names who approves and who maintains the agent, says where your data lives, tells you when not to build, and shows one published client result with how it was measured. Expect a written answer to each of the six questions below.

1. Will you write the job card before quoting? Ours: scope, success measure and approval points are agreed in writing before any build.
2. Who approves what leaves the firm? A person you name. Emails, proposals and posts wait for their yes.
3. Who runs the agent after launch? We do, every working day, with the numbers reviewed with you each month.
4. Where does our data live, and does it train a model? In your own installation in Germany, and no.
5. Will you tell us when an agent is the wrong answer? Yes: the Assessment can recommend a simpler process or no AI.
6. What do we keep if we stop? A full export of your data, rules and examples, then deletion on request.

Would you rather buy a tool and run it yourselves? Our comparison of ten AI agents for consulting firms covers that route, and the due-diligence checklist goes further before you sign.

At AI Jungle the price is published and set per firm, never per seat or per agent. The Leverage Assessment costs $999 / $2,999 / $9,999 depending on depth, credited in full if you continue. A build costs $15,000 once, then the Leverage Program runs the agents for $5,000 a month for the whole firm.

• Leverage Assessment: $999 Standard, $2,999 Department, $9,999 Onsite (travel billed at cost). Credited in full if you continue.
• Build: $15,000 once.
• Leverage Program, the same for every firm size: $5,000 a month for the whole firm, plus 5% of revenue growth or 15% of the gross-margin increase, never more than $15,000 in a month.

The build takes 30 to 45 days, and the Leverage Program carries a 90-day guarantee. Every term and condition is on the pricing page.