A lot, with very little

Sidenote was born with one idea: do a lot, with very little.

We are a small, senior team by choice, and we work with limited resources, also by choice. We like working in sync, on hard problems, and that is easier with two people than with twenty.

Today we have three products in production, StandupBot, SheetBest and agente+, and the team has not grown.

What changed in the last twelve months #

We adopted AI from day one. But it was in the last twelve months that the quality of the models jumped, and with it the list of things we can hand them.

For us that jump has a date: the launch of Opus 4.5. The code coming out of it was much better than anything we had until then, and from that point on we started delegating tasks that had not been worth delegating. Not because the model had learned to do something new, but because we started trusting what it did.

It was also in that period that we learned that there is a big difference between each person using these tools and the team using them.

One person with an assistant gets faster. Two people with two assistants both get faster, but producing different things.

Two examples of our own. To write documents and proposals, both people need the same rules on format and tone, or two different documents go out under the same signature. To plan marketing, the AI needs the same data and the same strategy, or each person gets a plan that is coherent with a different version of the company.

If one person’s version is out of date, incomplete or inconsistent with everyone else’s, you cannot trust what comes out. Keeping data, processes and systems coherent takes strategy and work. It does not come with the tool.

When that work is done #

When it is done, the value is of a different order. That is when you can say the organization has become AI-native.

The path, as we lived it, went like this. It starts with chat, where each person asks and pastes the answer. It moves to integrations, with AI inside the tools the team already uses. Then come workflows, where a sequence of steps runs on its own and a person reviews at the end.

After that come agents, systems that carry out a whole function, with defined goals and limits. And at the end, virtual “colleagues”, that the team relies on the way it relies on anyone else. We have not reached the top of this ladder yet.

For us, today, this means support for all three products is triaged, grouped and answered with automatic drafts, and we read everything that goes out. Code changes are small and frequent, with human review before every release, and release notes, documentation and product pages come from what changes in the code.

And there are things we do not automate: strategy decisions, design review and complex support cases.

What this is worth, and why it is still hard #

We know what this technology has given us. And we know most companies could get the same out of it, because there is nothing special about the processes we automated. They exist, under other names, in almost every company.

But for someone leading a team, a department or an SMB, it is still not easy to make this change without help. There is always something else to deal with. The technology changes month to month. You need to know where to start, how to fit the pieces of the different systems together, and how to keep them running after the first month’s enthusiasm wears off.

Why we launched services #

That is why we started offering services: we help leaders of teams, departments and SMBs in Portugal take this step. We always start with a process, and only then get to the technology.

We want to put what we learned building and using these systems every day to work for other companies, and see the number of companies in Portugal with these capabilities grow.

The workflows that keep StandupBot, SheetBest and agente+ running are the same ones we build for clients. And there are still two of us.