AUTOMATION AND AI
Less manual work. Decisions backed by data.
Process automation and artificial intelligence for small businesses: processes that no longer depend on someone copying, pasting and checking twice, and business information in one place from which you can get an answer.
Where the hours go in a small business
Manual work goes unseen because it is spread thin: each task costs little and is done every day.
- The same data is typed in two places
- It is noted down in one tool and then in the other. That doesn’t just use up hours: it brings errors with it, and they are found weeks later.
- The tools don’t talk to each other
- The ERP, invoicing and the warehouse each run on their own, and someone has to sit in the middle passing information from one to another.
- The report is put together by hand every time
- By the time the number is ready, the decision has already been made by eye. And so has the next one.
When AI adds value and when it isn’t needed
Artificial intelligence is one more tool and is justified like any other: by what it changes. These are the four questions we answer before putting it into a process, and if any of them has no answer, it doesn’t go in.
- What does it improve, and compared with what?
- If a simple rule, a better form or a mandatory field solves almost all of the problem, that is the way to go. It costs less, fails less and explains itself.
- What information does it see, and whose is it?
- A model processes real company data and sometimes its customers’ data. Where it is processed, what is stored and for how long are decided before anything is connected, not after.
- What does it cost each time it is used?
- Unlike software that is built once, here there is a per-use cost that grows with volume. We estimate it at the start so that it doesn’t turn up as a surprise on the third month’s invoice.
- Who reviews what it decides?
- Automating doesn’t shift responsibility to the machine. Where a mistake has consequences, the system proposes and a person approves.
What it costs to leave it
Hours of skilled people’s time spent moving information from one place to another, and decisions made by intuition because gathering the data costs more than deciding without it. It is a cost that appears on no invoice and that rises exactly when the volume of work rises, in other words when the business is doing well.
What is left running on its own
The processes that can be automated, running on their own, and business information in one place from which you can get an answer without assembling it by hand.
- Hours you no longer waste, and decisions backed by data.
- The systems you already have connected to each other, with nobody in the middle.
- Whatever is automated is monitored: you know when it fails, and who finds out.
What can be automated, and where AI comes in
- Process and workflow automation
- First we measure where the hours really go. We automate what is repetitive before what is showy; they are rarely the same.
- Connecting your business management, invoicing, warehouse or CRM systems
- So that what you already have no longer needs a person passing data from one side to the other.
- Artificial intelligence applied to specific tasks
- Classifying, extracting information from documents, preparing drafts or ordering a work queue. With a person reviewing wherever the decision matters.
- Organised data to base decisions on
- One place where the same number is the same whoever looks at it, and a dashboard that answers the questions people really ask in meetings.
How we work
The same method as on all our projects, applied to a process that is already running and can’t be stopped while it is being changed: we start with a preliminary analysis that measures where the hours go.
- UnderstandWe measure the real process: how many times a day, how long it takes and who does it.
- DefineWe produce a signed list of what is automated and what stays manual.
- DesignWe decide which system is the source of truth for each piece of data before connecting any.
- BuildShort releases: one complete process working before the next one starts.
- LaunchWe run it alongside the manual way until the two results match.
- EvolveWe monitor whatever is automated: a process with no oversight is another problem.
How it is built inside
An automation that nobody watches is a silent breakdown. What decides whether this holds up is not which tool is used, but what happens the day one of the systems in the middle stops responding, because it will.
No two operations are alike. What doesn’t change is that these responsibilities are kept separate:
- Interface
- the dashboards and forms where a person reviews, corrects or approves
- Integrations
- the contract with each tool you already use, and what happens if it fails
- Business rules
- when each automation fires and under what conditions, written down in one place
- Data
- which system is the source of truth for each piece of data and what is kept from each step
- Each integration knows what to do when the other side doesn’t respond: retry, alert or stop.
- A record is kept of what ran and with what result, so that a number that doesn’t add up can be checked without guesswork.
- The code we develop specifically for your project is yours, with its repository and documentation, once the amounts due have been paid. If we use third-party APIs, libraries or services, their licences are identified before we incorporate them.
Before you write to us.
Which processes in a small business can be automated with AI?
The ones that repeat, follow clear rules and leave a trail: moving invoices and orders into your system, sorting documents or emails, preparing drafts and putting reports together. AI comes in where there is text or documents to read, always with a person reviewing the result; where a simple rule is enough, it is not needed. To choose where to start, there is the guide What to automate first in a small business and, if you start from a spreadsheet, Automate invoices from Excel.
Can supplier invoices be automated?
Yes, and it is one of the automations you notice soonest. An incoming invoice is read, checked against the order and passed to accounting without anyone retyping it, with a person reviewing whatever doesn’t match. What isn’t automated blindly is anything with tax consequences: that data is checked before the automation uses it. The rules on invoicing systems, including Verifactu and IGIC, have their own guide: Verifactu and IGIC: what Canary Islands businesses should know. And to choose a route from a spreadsheet, automate invoices from Excel: four routes.
Can orders that arrive by WhatsApp or email be automated?
Yes, it is among the most common cases. The message is read, customer, product and quantity are identified, and the order is created in your system for a person to confirm before it counts. Two conditions are checked before starting: WhatsApp’s terms of use and consent requirements, and what happens when a message is ambiguous; what the law requires of a chatbot is explained in the guide WhatsApp chatbot: what the law requires. Where prices or stock change often, the system proposes and a person decides.
What is AI consulting for small businesses, and when is it worth it?
It means deciding, before building, whether artificial intelligence adds anything to a specific process. What it improves and compared with what, what information it sees, what each use costs and who reviews what it decides. It is worth it when there is a repeated task involving text or documents and a person can review the result; it isn’t when a simple rule or a better form solves most of the problem. That comparison is the preliminary analysis: it is made before proposing anything, and sometimes the answer is that AI isn’t needed.
Do we have to change the tools we already use?
No, that is not the starting point. The first step is to connect what already works, because replacing a tool the team has mastered carries a learning cost that is almost never counted in the budget. We propose changing something only when that tool is precisely what prevents automation, and then we say why and what is gained.
What happens to our data if artificial intelligence is used?
We decide it before we start, and in writing: what information goes into the model, where it is processed, what is stored and for how long. When the data is sensitive, the first option is not to take it out of the system, or to take it out without identifying anyone. And if a provider doesn’t allow you to check how it handles that information, it is ruled out before it is integrated, not after.
Does this replace part of the team?
That is for the company to decide, not the tool, and it is best decided before starting. What we do is remove the work that adds no judgement (copying, pasting, checking twice) so that hours are spent where judgement is needed. Where a decision has consequences, a person still approves, and that is not a concession: it is how it is designed.
And if the automation gets it wrong?
It is designed so that you notice. Each automated process keeps a record of what it did, raises an alert when something doesn’t fit and retains a manual way to carry on while it is fixed. An automation that fails silently does more harm than the manual work it came to replace, which is why monitoring is in scope from day one.
We start with a preliminary analysis: scope, deliverable and fee agreed in writing before we begin. We usually reply in under 24 working hours.
If what you have built already works and what is missing is maintaining it: technology evolution.
To decide where to start automating: what to automate first in a small business.
If you automate invoicing and work in the Canary Islands: Verifactu and IGIC for Canary Islands businesses.