How to build a second brain for your company. The first one stays with its founders, nobody is after it
The last time I applied for a job as a deputy director, they reassured me right at the door: “Automation is fine here. We have a CRM.”
A month later I was drowning. Not in work. In the amount of information I had to hold in my head at the same time, analyse, and then somehow use to make management decisions.
“We have a CRM” meant exactly one thing: we have a place where the sales team logs their calls. How that was connected to the real problems of the company, nobody could say. Welcome to the reality of management automation.
That is what this article is about. It is written by a manager, not an IT person: twenty-five years on the owner’s side, five years in software development as a client, and three years in AI as an active user and builder of all sorts of things for my own businesses, and now for several clients too. I will briefly show how today you can build a second brain for a company. One that pulls in data, remembers it, connects it, analyses it and comes to you with a report on its own. There will be a little history (electricity and Kodak), the prices of the big players, and what I have already tested on myself.
And one warning about the style. Business articles are usually written for very busy and very smart men who do not like to show that they do not know something. This one is written by a woman who for many years was the only woman in the meeting room. For example, in Cyprus women make up 8.5% of directors of the largest listed companies, and in Europe only Hungary is lower (EIGE, 2025). In those meeting rooms I learned one rule: if you want people to listen, speak simply but vividly. To the point, but never boring.
So I will keep that style: few clever words, many facts, and plenty of examples from my own experience.
Thirty programs and one head
That CRM, by the way, was a good one. The problem was not the CRM.
Today there are so many programs for managing tasks, projects, sales and stock that none of them can surprise anyone. Everyone has a CRM. Everyone has spreadsheets. Accounting, telephony, email, messengers, ad accounts, a meeting tool. A mid-sized business easily collects twenty or thirty of them.
The problem is that they hardly talk to each other. They are connected by one single integration: the head of a manager-analyst. I know what I am talking about, because for twenty-five years that head was mine. Finance, accounting (yes, these are different things), legal, methodology, IT, sales and marketing, and if there is production too… then you only have enough energy to put together, once a month and somehow, a picture of the business for the owner from all these puzzle pieces.
Sadly, a head is not made of rubber. It cannot hold hundreds of parameters at once. And comparing them with what was happening six months or a year ago at the same time, but with slightly different starting conditions, is a job for a very calm person with a very large supply of coffee.
For twenty-five years the industry tried to cure this with dashboards. Analysts themselves admit the result: about 19% of employees used BI systems regularly in 1998, and about 25% in 2022 (TechTarget). The tools became prettier, but the number of people they actually help hardly grew.
A dashboard shows what happened. The owner asks why. And what to do about it now.
For the minutes: nearly thirty programs have been implemented, but there is only one integrator, and that integrator is a human being who gets tired.
Why electricity did not make factories rich right away
By the end of the 19th century electricity was everywhere there was money: light bulbs, power stations in New York and London. Factories happily put an electric motor where the steam engine used to be. And nothing happened. Productivity barely grew. In 1900, electric motors provided less than 5% of the mechanical power in American factories (Tim Harford, BBC).
The jump came only in the 1920s, almost forty years later. And it was not the motor that made it. It was a new factory layout: single-storey buildings, production lines that follow the flow of materials, a small motor at every workstation, new training for people. The economist Paul David called this the “dynamo paradox” (Harford, FT).
A hundred years later we are doing the same thing. We buy AI and put it where the old steam engine stood. The same mess in the data, the same thirty unconnected programs, only now with a chat on top that answers very confidently.
Anthropic, the company that makes Claude, says it plainly in its research on how businesses use AI: on complex tasks, adoption is limited “more by access to information than by the capabilities of the model” (Anthropic Economic Index). Gartner predicts that by the end of 2026 companies will abandon 60% of AI projects that do not have AI-ready data (Gartner).
For the minutes: first rebuild the flows, and only then install a new modern motor.
What has changed in three years
Five years ago I dived into software development deeper than I wanted to: first as a client, then as an analyst, then as a tester, and then as a product manager… yes, yes, that is called a founder. And three years ago I dived into AI tools as an active user. AI arrived just in time, I was simply no longer coping. But…
At first it was like Alice and the Red Queen: you run as fast as you can just to stay in the same place. Every month a new tool, a new “killer of everything”, a new subscription. And the realisation that your product, the one you had been building for three years, would very soon be of no use to anyone in the form you built it…
Then it became clear what had really changed. It was not another task-management program. There are enough of those. What appeared was the chance to build simple solutions for yourself quickly, with the help of AI development. And above all, they must be simple for the people who use them. They pull in data on their own, internal and external, sort it and connect it.
I also realised that now you can set up AI analysts by area: finance, legal, marketing. Some work as executors. Others work as heads of department: they collect the executors’ work and report to me. Almost like a real staff, only without sick days and without “I thought this was not my area of responsibility”.
No, and again no, I have no illusion that AI will fully replace human work. But it takes that work to a new level. It is like with assembly lines: you can push dolls’ heads onto their bodies all day (a childhood memory from some film just came back…), or you can be the engineer who sets up the robotic machines that now do it.
It is also like what happened to bookkeepers after spreadsheets appeared in 1979. The clerks who added up columns were replaced by the program. But the number of accountants and analysts grew, because people started expecting more from them: not just to add up numbers, but to work out “what happens if we do it differently” (NPR Planet Money). The profession itself did not disappear. The questions became harder, and the demands on people became higher.
And one detail I personally like. Harvard Business School pulled together studies from about a hundred countries: women use generative AI about 22% less than men, and the gap is hardly closing (Poets&Quants on the HBS research). So a woman who builds business management on AI is still a rarity squared. I will take that as a compliment.
For the minutes: what is new is not the idea of automation, but the fact that now you really can build it for yourself, even on your own.
The brief arrives before the coffee
From theory to practice: I promised facts. This is how my task tracker works.
During the day or in the evening I dictate the current situation into a chat bot: what is done, what I am planning, what worries me. After a meeting I drop its transcript into the same chat. No cards, no “fill in the field, choose the status, assign the person”.
In the morning, before the coffee, the brief arrives. What is on fire, what can wait, who was supposed to do what and did not. And also in plain words, in the same chat, I set up exactly what the report should look like. For example, I get several of them, a separate one for each project. That is easier for me.
Need to look at one department or one specific employee? I just say it to the same bot: “show me what is happening with sales over the last two weeks and which sales managers are behind”. The report comes in the form that is easiest for me to read.
Marketing works the same way. AI collects what people are really searching for, and SEO promotion is built on that. Content is planned on statistics, not on the inspiration of a marketer who felt like pretty pictures today.
The bot is also a team bot: each person in the team has their own level of access to information and to the reports they can request. It can also prepare meeting agendas, and also… and also a lot more, because I set it up exactly the way that is convenient for me and my team.
For the minutes: the best way to enter information is simply to say it out loud. Everyone is very tired of forms to fill in. There are enough forms at banks and government offices already.
The big players are already there. So are the big prices
If you think these are the dreams of small enthusiasts, look where the big companies are going.
Microsoft sells Copilot on top of a company’s email and files for roughly $21–30 per user per month, and agents are counted separately, in credits (Copilot pricing overview). Salesforce charges for its AI agents $2 per conversation or $125 per user on top of the licence (Salesforce). Glean, a company-wide search across all business systems, does not publish its price at all: “request a demo”, and according to reviews it starts at 100 seats and from $50,000 a year (GoSearch). Palantir deals start, according to analysts, at around a million.
Morgan Stanley gave its advisors an assistant that answers from the bank’s internal library, and 98% of the teams use it (CDO Magazine). It worked because first there was a checked and organised knowledge base, and only then a chat.
Notice one detail. Copilot searches inside Microsoft. Salesforce agents work inside Salesforce. Each big player sells AI on top of its own system and quietly assumes that your data already lives there and is already in order.
As a result, small and mid-sized businesses get two extremes. Either a cheap general chat that, when asked “how much do we owe our suppliers”, honestly says it does not know. Or an enterprise sales cycle with demos, tenders and contracts, for which a twenty-person company simply has neither the money nor the people. The middle is almost empty.
In May 2026 Anthropic became the first big player to enter this middle, with Claude for Small Business: ready-made processes for accounting, sales and invoices on top of QuickBooks, HubSpot and Stripe (Anthropic). A good sign. But it also expects someone to connect your systems, put your data in order and define which questions the owner actually asks.
This middle is exactly what interests me. Simple solutions for small and mid-sized businesses, with no tenders.
For the minutes: the big players sell AI to big companies. The middle, and even more so small business, is still open.
Automated chaos costs more than ordinary chaos
Let me be honest right away: what I do is not unique. Many people offer something similar today. But most often they are IT people who once had some management experience. With me it is exactly the other way round. I am a manager with long experience who knows where information usually gets lost and why a business starts to sink. First I set everything up on myself and my own projects, experiment, make my own mistakes. Only then do I set it up for a client.
This is more a warning than an advert. Because there are two honest paths.
The first: do it yourself. If you can build an internal team or are ready to learn it personally, today you have everything you need. There is a sea of open information, and courses are cheap. Take it and apply it to your current tasks, not to some abstract “digitalisation”.
The second: professionals. Then look at their background. Already from the first questionnaire it should be clear that you are dealing with a team led by a manager. If they mainly ask which programs you have installed, that is a bad sign. If they ask how money, requests and tasks move in your company, who is responsible for what, and where information gets lost along the way, you are in the right place. First you build the information flows, then you automate them. The other way round you get automated chaos. It is faster than ordinary chaos and costs more.
The price of a mistake is already known. Air Canada’s chat bot promised a passenger a refund that was not in the rules, and the tribunal decided that the company is responsible for what its bot says (Moffatt v. Air Canada). Deloitte delivered a report worth 440 thousand dollars to an Australian government department with an invented court quote and returned part of the fee (CFO Dive).
And yes, such systems cannot be built without professional IT people. Connecting systems, setting up access, making sure everything works every day and does not crash at night: that is their job, and they do it better than any manager. The only question is who leads. An IT person will sell you software. A manager will first map your processes, to understand what exactly should be automated.
For the minutes: ask who in their team will map your processes. If the answer is nobody, run.
One text box
In a little over two years, ChatGPT taught almost a billion people one simple habit: there is one box, you write in it in plain words, you get an answer. 100 million people a week in 2023, 900 million at the start of 2026 (TechCrunch).
My deep conviction is that a modern product should aim for exactly this. One box on top. Everything else under the hood.
Microsoft CEO Satya Nadella says almost the same, only in engineering language: business applications are basically databases with business logic, and that logic will move to AI agents (Outlook Business).
There are honest critics too. Designer Amelia Wattenberger wrote that chat bots are “a terrible interface”, because a good tool shows you how to use it (Wattenberger). She is right about one thing: an empty box is scary if you do not know what to ask. So the real work is to make sure that everything the owner may want to ask about is already organised under the hood.
That is exactly why I am now going to rewrite one of my platforms, FindPro, from the ground up. What was good four years ago, when I came up with it, is of no use to anyone today. Forms, filters, long questionnaires, personal accounts. Does it hurt? A little. But it was these failures that showed where to go and which new hypotheses to test.
For the minutes: today’s reality is simple: if people have to learn which buttons to press in your product, it is out of date.
What is on my desk right now
No selling, just so you know what I am busy with.
- Rewriting FindPro to make finding professionals in Cyprus truly simple and convenient.
- Building a system of templates and AI analysts for management, first on my own projects, then for clients from small and mid-sized businesses.
- Every morning I read my brief and grumble that the list of tasks only grows: so much new and interesting appears every day that I want to try it with my own hands and see how useful it can be for my tasks.
And a question for you. Where in your business does information get lost most often? If someone came to you and said, “Let us automate something for you…” What would it be? Which business process would you hand over to automation first? Write to me. I collect these places. They are the best management textbook I have ever seen.
Svetlana Siakalli. Co-founder and operating CEO of FindPro and iPRO WAYs, founder of LensDeck. More than 25 years in management on the owner’s side: finance, IT, sales, marketing.





0 Comments