The New AI Employee
Guide

AI Automation Consulting: What Happens Before Anything Is Built

AI automation consulting starts by reading how your work is done today. What the review covers, how jobs are chosen and what the proposal must say.

The New AI Employee6 min read
A consultant's notebook mapping office routines as boxes and arrows, beside a laptop showing read-only folders and a magnifying glass.
Short answer

Good AI automation consulting starts by reading how the work is done today, in your emails, files and records, before proposing anything. You end up with a short list of jobs worth automating, what each robot will draft, who approves it and one plain monthly price per job.

What does an AI automation consultant actually do?

An AI automation consultant studies how your office work is done today and tells you, in writing, which parts are worth automating, how, and at what price.

AI automation consulting sits between a first chat and a build. Someone from outside looks closely at your routine work and says what a robot could prepare, what must stay with people and what isn't worth automating at all. Business process automation consulting and workflow automation consulting mean the same stage. So do many offers sold as AI consulting services.

Done well, it's mostly reading and asking. A consultant who shows up with a ready answer before seeing any of your work is selling a product, not advising. Done badly, it ends with a slide deck full of ideas and no clear first step.

Ours starts with a plain answer, not a pitch. Three lines are enough: what you do, which tasks eat your time and roughly how many hours a week they cost. From that we can usually say whether an AI employee fits at all, before anyone opens a single file.

What is the read-only review?

The core of AI automation consulting is a read-only review: looking at real emails, files and records to see how the work actually happens, not how people say it happens.

People describe their routines from memory, and memory smooths things out. The real trail shows the odd cases. Say a small wholesaler has a supplier who sends invoices as phone photos, a customer who orders by replying to an old thread, and a spreadsheet only one person understands. An automation has to handle exactly those, so we find them before designing anything.

The review runs only with your permission, on the accounts you choose. Nothing is created or changed. Here's what we typically look at.

  • The mailboxes where work arrives: orders, supplier invoices, requests, questions.
  • The folders where documents are filed, and how tidily.
  • The sheets and lists people keep by hand.
  • The accounting records: how invoices and bills are entered, coded and attached.
  • Who handles each step today, and who signs off the result.

How do you choose the jobs worth automating?

The best jobs repeat often, follow rules you can write down, and end in a document, an entry or an email that someone can check quickly.

Every business has more candidates than it should start with. The review usually turns up a long list, and the real work is cutting it down. A job that happens rarely, or needs a judgment call every time, is real work but a poor candidate. A job that happens daily by the same rules, and that someone types by hand, is usually a good one.

This is also where each job gets its size. Counting how many orders, bills or requests arrived over a few months, instead of guessing, turns a vague wish into a job that can be priced. A bakery that gets forty orders a week by email is a clearer case than one that says "quite a lot".

Every job we build follows the same path. The work is read from email, files or sheets. It's put into one clean record and checked against what the business already knows. Rules decide where it goes. A draft is prepared, and it's handed over complete with its source beside it. Laying a candidate job over that path shows quickly if it fits.

The table shows how candidates usually sort out.

SignalGood candidatePoor candidate
How often it happensDaily or weeklyA few times a year
The rulesFit on one pageDifferent every time
Where the work arrivesOne predictable placeAll over, often by phone
The resultA document, an entry or an emailA decision or a negotiation
Checking itFaster than doing it by handAs slow as doing it by hand

What gets decided for each job?

For each chosen job, we settle exactly what the robot will draft, who approves it and what it must never do.

This is the part most often skipped, and it decides whether the automation can be trusted. Our rules are fixed: every step starts with your approval, then you choose when it goes automatic, every action is logged and one command stops any robot. What's left is the detail, job by job.

Think of a plumber's office and the supplier bills. These are the questions we'd answer for that one job.

  1. Where does the work arrive, and what starts the robot: a new email, a file in a folder or a row marked ready?
  2. What exactly does the robot prepare: a draft invoice, a draft bill, a draft email, a row in a list?
  3. Where does the draft wait, and who gets told?
  4. Who approves it, and who steps in when that person is away?
  5. Which cases go to a person instead of being guessed, like unknown suppliers, refunds or missing data?
  6. What must the robot not do on its own: send, pay, delete or add a new supplier or customer?
Interactive

Are you ready for a consulting review?

Tick what you already have. Every line you leave blank is something to do before the first chat.

What should the written proposal say?

It should be short and say four things: the jobs worth automating, what each robot will draft, who approves it, and one plain monthly price per job.

This is what AI automation consulting should be judged on. A proposal you can't sum up in a few lines is a warning sign.

Ours lists the jobs in the order they'd be built and describes each in plain words. It states the approvals and limits above and gives one monthly price per workflow. The price depends on the hours the job takes off your plate and how complex it is. A job that needs several AI employees working together costs more than a simple one.

It also says how the service runs. The AI employees are rented monthly. We build them, run them on our servers against your own accounts and look after them. There's no one-off sale. If the rental stops, the workflow stops, and every record stays in your own mailbox, folders and sheets.

How long will it take? It depends on the work. Simple tasks are ready quickly, complex ones take longer.

What if the answer is no?

Honest AI automation consulting sometimes ends with advice not to automate a job, or not yet.

There are good reasons. The job may happen too rarely to justify a monthly rental. The rules may live in one person's head and need writing down first. The work may arrive in so many forms that the first step is gathering it into one mailbox or folder, which you can do yourself. Or the job may be mostly judgment with a small mechanical part.

A no is still useful. It often comes with a smaller step that makes a later yes possible, like a shared list kept by hand for a month, or one address for supplier invoices. It also tells you something about the adviser. A consultant who never says no is running a sales process.

What should you prepare?

You can make the review faster and sharper by preparing a few things before the first chat.

None of it needs polish. A rough note, a few real examples and a clear idea of who approves what beat a long description. The readiness check below lists the basics. Tick what you have, and the result tells you what to prepare next.

The most useful thing is your own count: how often each job happens and how long it takes, by your own rule. That number, in hours, is what an AI employee is built to give back. We promise nothing about how much any automation will recover. The review is how both sides find out.

Questions

What does an AI automation consultant do?

They study how your office work is done today, ideally by reading real emails, files and records, and tell you in writing which jobs are worth automating, how each would work and what it would cost.

What's the difference between AI consulting services and building automation?

Consulting decides what to build and why. Building makes it work. With us the two go together: the review leads to a proposal, and the same team builds and runs what you agree.

Do you need access to our systems for the review?

Read-only access to the accounts you choose, and only with your permission. Nothing is created or changed during the review.

How long does an automation project take?

It depends on the work. Simple tasks are ready quickly. Complex ones, with several AI employees working together, take longer. The proposal tells you what applies to yours.

How is the price set?

One monthly price per workflow, stated in the proposal. It depends on the hours the job takes off your plate and how complex the automation is.

Tell us your workflow. We will build it around you.

Write to hello@thenewaiemployee.com in three lines: what you do, which tasks eat your time and roughly how many hours a week they cost. You'll get a plain answer.

Talk to us Related service: Services

Keep reading

Write to us

Tell us which hours you want back.

Three lines are enough: what you do, what takes your time, how many hours a week it costs. Write to hello@thenewaiemployee.com.

Offices in Sheridan (Wyoming), Panama City, London, Bucharest and Milan.