NACH Analytics

The job you redo every week

AI automation for small business, one task at a time. The report you rebuild every month goes to software that doesn't get bored or skip a step, and that tells you when it stops.

Our AI assistant asks about your project and can build you a free plan.

The problem

Someone in your business does the same task every week, the same way, from the same files. They export two reports, paste them into a spreadsheet, fix the columns that never line up and email the result to four people. It takes half a day each time, and it has taken half a day for three years.

Nobody has fixed it because it isn't urgent. It costs money without anyone noticing, and it's fragile in a way that only shows when the person who does it is away or leaves, and it turns out the steps only ever existed in their head.

How we approach it

First we watch the person who does the task do it, including the steps they've stopped noticing. Those steps hold the logic a requirements document misses: the exception for one customer, the column that has to be fixed by hand, the check that stops a bad number reaching an invoice.

Then we automate the whole path, start to finish. A task that still needs someone to kick it off and tidy up afterwards has only been shortened.

Two things ship with it, so you can trust it without babysitting it:

  • A log of every run: what it did, what it saw, whether it worked. If you can't see what it did, you'll check it by hand, and then you still have the job.
  • A run guide: how to tell it worked, and what to do when it stops. It will stop one day, when a system changes, a password expires or a file arrives in a new format.

We use AI where it helps, like reading documents or pulling fields out of unstructured text, and we spell out where it's allowed to be wrong and what catches the mistake. Where a simple rule is more reliable, we use the rule, because a language model in the middle of arithmetic makes a simple job unpredictable.

Failing loudly

An automation that fails silently is worse than the manual task it replaced, because people stop checking. So by default it stops and tells someone: a Monday report two hours late does less harm than one that arrives on time and wrong.

Before we automate anything, we break the job into its smallest repeated steps. The big task is rarely automatable as a whole. The small steps inside it almost always are, and that is where the hours hide.

A principle we work by
A monthly client report that builds itself and waits for a person to press send.

A worked example

Take a services firm that sends each client a monthly report. Someone exports hours from the time-tracking tool and invoices from accounting, matches them by client, works out the write-off, formats the file and sends it. It takes two people most of a day, every month.

Automated, it runs on the first working day. It pulls both sources, matches them, flags the clients whose write-off moved more than usual (three, in the figure), builds the file and sends it for approval. The report doesn't go out on its own: a person still presses send, because a wrong client report is expensive and a late one isn't.

The log shows every run and every flag raised. If one month the time-tracking export changes format, the run stops at 6am and says so before it can produce a wrong number, and the run guide says exactly what to check.

Fixed scope

Automation Sprint

What you get

  • One recurring task automated end to end: a report that builds itself, data moved between two systems, or documents read and summarized with AI
  • Logging, so you can see every run and what it did
  • A run guide covering how to check it, and what to do when it stops

What we need from you

  • A walkthrough of how the task is done today, by the person who does it
  • Access to the systems involved, or a test account
  • A decision about what should happen when the task fails

What is out of scope

  • More than one process
  • Custom applications
  • Integrating a system that has no API. If no program can connect to it, we'll tell you before you spend anything
How long it takes
About three weeks. Start dates depend on current availability.
What happens after
Thirty days of fixes are included. The same optional care plan applies, and for automations it matters more: something that runs every week needs someone watching it.
Price
We give you a fixed price before we start, usually within two business days.

Need something bigger?

Most businesses have five of these tasks. We still recommend starting with one: automating something small and watching it run for a month is the cheapest way to learn how your systems behave.

If a process needs same-hour intervention when it breaks, tell us early, and we'll tell you whether we're the right people.

Every project is handled directly by a member of our team.

If you let us publish what we build together, you get a discount. The work has to be something we can publish, and you approve every word before it goes live. Our team applies the discount when we quote.

Questions people ask

What kinds of tasks are worth automating first?

Frequent ones that follow fixed rules and that people dread. A monthly task that takes a day is usually a better first candidate than a daily task that takes five minutes: the payback is easier to see, and if it fails, it's less urgent.

What if the software we use has no API?

Then we tell you before you spend anything. There are usually workarounds, like scheduled exports or file drops, but some are fragile, and a fragile automation is a future outage with your name on it. We'd rather lose the work than sell you one.

Will this replace someone's job?

In a small business, almost never. It takes away the part of the job nobody wanted. It's worth talking that through with the person who does the job before we start.

Can you automate reports we build in Excel or Google Sheets?

Yes, and it's often a good first project. If the file comes from an export and the steps are the same each time, it's a good candidate. If the person doing it makes judgement calls halfway through, we map those out first, because that's the part that usually gets missed.

Who fixes it when it breaks?

Thirty days of fixes are included. After that, you can take the optional care plan or contact us when something comes up. The run guide is written so a capable person on your side can handle the common cases without us.

Do you use AI in these?

Where it earns its place, like reading documents or handling messy text. We don't use it for arithmetic, or anywhere it would hide what happened. Every run is logged either way.

Which task would you hand over first?

Tell us which one takes up the most of your time. If we can't help, we'll say so.

Our AI assistant asks about your project and can build you a free plan.

Or book a 30-minute call

Nacho AI

NACH Analytics' AI assistant

Book a call