Sort the work before you buy the tool
Nihul Consulting puts AI on the routine work that repeats: the inbox, the quote you have written forty times, the report somebody rebuilds every Friday. The tool does a fast first pass and a person checks the result before it leaves the building. The judgment work stays where it is.
What this covers
-
Which tasks are safe to paste
Write down the ten things you would most like off your hands, then mark which ones contain a client identifier. That afternoon settles most of the tooling argument before you have looked at a single product.
-
Back-office automation
The repetitive admin comes off the team's plate.
-
The prompts your team reuses
Templates kept where everyone can find them, so the fiftieth draft in ChatGPT or Claude reads like the first one you approved.
-
Training so the setup survives its first month
A tool nobody was shown how to use gets quietly abandoned. We sit with the people who will actually run it until they trust what comes back.
-
Data rules agreed in writing
What a tool may see, and what is kept out of it entirely, settled before anything is switched on. The free tiers of the big chat tools say in their own terms not to put sensitive or confidential information into them, which is the vendor telling you which bucket its free product belongs in.
Everything else an engagement can draw on is in the full list of services.
Where the two hours a week go
A few routine tasks quietly take a couple of hours out of every week. Most can be set up so the tool does the first pass in minutes and a person checks the result before it goes out. These are the ones that come up most often.
-
The inbox, sorted before you open it
Mail grouped by what needs a real answer, with the routine replies already drafted in your own wording. You keep the send button, and anything sensitive stays with a person.
-
Quotes drafted from the ones you already sent
Most new quotes are variations on your own history. The draft arrives with the right line items and your usual terms; you set the price.
-
Chasing unpaid invoices
Polite, on-time reminders in wording you approved, so nobody keeps a running list in their head. A first gentle nudge is usually enough.
-
Meeting notes into a list of who owes what
A rough set of notes becomes an action list you can circulate the same day, while everyone still remembers the meeting. Someone confirms it before it goes out, because a misheard commitment written down is worse than none at all.
-
The weekly report
The same short summary every week, assembled from figures that live in three different places, ready to read on Monday morning.
-
Typing the same details three times
A new order or contact is entered once and carried into the CRM and the accounting tool. Worth building slowly, because a quiet copying error spreads everywhere at once.
-
The questions that arrive every week
Opening hours, pricing, what happens next. Answers drafted from your own notes and past replies, so the wording stays consistent.
-
A forty-message thread, read down to the decision
A long contract or a dense report cut to the few points that need a call, with a link back to the source so you can check the detail.
Most of what a small business does all day is safe to put in an ordinary cloud tool, and the part that is not is smaller than people fear. What decides it is the text you would paste in. That makes it a sorting job on your own work, and the question of where the servers live mostly answers itself afterwards. Running open models on a machine in your own office is right for a narrow set of businesses: the ones whose contracts forbid outside processing, and the ones that already employ somebody who patches servers. Everyone else buys a machine and inherits a second job.
Start a conversationCommon questions
What does AI enablement mean in practice?
It starts with sorting your own work. Write down the ten things you would most like off your hands and mark which ones contain a client identifier. That sort tells you which tasks a tool can take, and then we build the handful of setups the business will rely on every day.
Which AI tools do you work with?
Mostly Claude and ChatGPT, plus whatever already fits the tools you pay for, and locally run open models where a contract or an obligation requires it. We keep the setup portable so it can move when a better option appears.
Will AI replace my staff?
That is not the aim. AI is good at the repetitive work that wears people down, so it clears that off their plate and leaves them the judgment work only people do well. Used this way it makes a small team steadier.
How is my business data handled?
Nothing goes into any tool until we agree what it may see and what is kept out of it entirely. Those rules are set in writing before anything is switched on, so the team knows where the lines are. The free tiers of the big chat tools tell you in their own terms not to submit sensitive or confidential information, and that is the tier most people sign up on.
Which everyday tasks can AI actually take over?
The routine, repeatable ones: sorting the inbox and drafting replies, building quotes from past ones, sending invoice reminders, turning meeting notes into action lists, assembling the weekly report, moving data between your tools, answering repeat questions, and summarizing long documents. The tool does the first pass in minutes and a person checks it before it goes out. Anything that turns on judgment or a difficult relationship stays with your team.