AI that makes it into daily work

I help organisations put AI into the processes they already run. I help choose the right solution and, where needed, take part in implementation.

Book a call Describe one process Describe one process and I'll think it through with you

Atrium of a modern office building in daylight.

When to get in touch

Many organisations have tried AI, but it has not reached day-to-day work. Typical situations:

Copilot licences have been bought, but usage is sporadic.
People use AI tools on their own, and no one knows what data goes into them.
A choice has to be made among several offers or tools, and you don't know which one fits your process.
The same manual work repeats every day across several systems.
Leadership expects results from AI, but there is no clear owner or first use case.

If any of these is familiar, I start with one process and take it through to the end.

Why do AI pilots so often stay out of daily work?

A tool alone does not change the work. These four things decide whether a pilot stays a trial:

Data and permissions are messy

If nobody knows which files the tool can see, nobody will trust it. Permissions have to be fixed before a wider rollout.

Decision rights are undefined

There is no agreement on which steps a person confirms and which the system may take on its own. So the work stays a trial.

A purchase does not change the work

Buying a licence or a model does not change a workflow. Change comes when the tool is tied to a real task, real data and a named owner.

Benefit is not measured

If the result is not measured before and after, you cannot decide whether to continue, expand or stop.

Data and control

Data location

I choose a setup where it's clear where data is processed: inside your Microsoft 365 tenant, in an EU cloud, or on your own server if we use an open model.

Permissions before Copilot

Before rollout I review who can access what, so AI doesn't surface documents people shouldn't see.

Agents act within agreed limits

Each agent has defined tools and permissions; consequential steps are approved by a person, and actions are logged.

GDPR and the EU AI Act

I consider data protection and the EU AI Act from the design stage, not afterwards.

Services

I start where you have the most manual work or uncertainty, and I take it through to the end.

AI and process consulting

I map processes, data and risk, and leave you with a clear recommendation on where to start.

A review of your workflows and data sources
Risks, including an initial assessment of EU AI Act requirements (not legal advice)
Three use cases with an initial ROI estimate

Integrations

I connect AI, Copilot and agents to the systems you already have, so the work stays where it already runs.

Integration with Microsoft 365, CRM or ERP
One workflow wired from start to finish
Measurement before and after

Solution selection and requirements

I help you define what the solution must do, write the requirements, and compare offers and tools.

Requirements spec
Evaluation criteria
Offer comparison and a recommendation with risks

Website AI readiness

I assess whether your site can be found in AI search, is technically sound, and is usable by agents.

AI search (GEO/AEO): ChatGPT, Gemini, Copilot, Perplexity
Technical SEO and structured data
Agent interfaces: MCP, A2A, llms.txt, APIs

AI in your processes: I help put AI into the processes you already run. The aim is that one specific task starts working in daily use.

Setting up assistants for daily use: Grok, Claude, Manus, Meta AI, Gemini and Microsoft 365 Copilot
Your own bots and assistants on the channels you already use: email, Slack or Teams, the web and chat tools
Workflow integration: I connect AI to the systems where your work already lives, including cloud AI services when those are part of your setup.

AI services on AWS, Azure and other clouds
Agents: I help you choose and put into use an agent that fits the job.

Custom-built agents
Open-source frameworks, such as Hermes Agent and OpenClaw
Managed agent platforms, such as Microsoft Copilot Studio
Open-weight and closed models: I help you decide whether an open-weight or a closed model fits the work. I compare privacy, cost, control and quality against your constraints.

Open-weight models: more control and the option to run them on your own server
Closed models: a ready-made service, with a different cost and data model
Keeping it running: Once the solution is in production, you need to see how the model and the agent behave.

Model monitoring and logging
Observability of agents in production
LLMOps: usage, quality and incidents

How I work

1

Map

I establish which tasks take time, what data they use and what risk they carry.
2

Select

I compare tools, vendors and architecture and pick one or two use cases with a measurable benefit.
3

Integrate

I connect the solution to your systems. I put permissions, guidance and owners in place before the circle of users grows.
4

Handover

You receive the playbook, the measures and an agreement on who owns the solution. The solution stays with you and does not depend on me.

Things I've built

Below are products I built myself and made publicly available.

ted-mcp.eu

An open connector that links AI assistants to EU tender data.

Agentic Gov

My personal project for ideas and experiments: how an agent works within a human decision and mandate.

I have more than 20 years of building and integrating systems.

About me

I am an independent AI expert and fractional CTO. I put AI into processes that already exist.

20+ years of software development and systems integration
Products built on public-procurement data
I write about AI and accountability at alvarlaigna.com

More about me and the company

Before you get in touch

What does it cost?

My hourly rate is €360. On longer projects a lower day rate applies, around €600. I send a precise proposal after the first conversation.

How long does it take?

A single-process solution usually takes a few weeks; more systems and more complex logic take longer. We agree the timeline before work starts.

Do we need our own IT team?

No. You need someone who knows the process and access to the relevant systems. I handle the technical side.

Who actually does the work?

You work with me directly. I map, build and hand over the solution myself.

Which systems can AI connect to?

The ones where your work information already lives: Microsoft 365, CRM, ERP, accounting software, document management, databases, public APIs, AI services on AWS and Azure, and channels such as email, Slack or Teams, the web and chat tools. Before building I check which integrations are needed and safe.

I'm not technical. Is that a problem?

No. Your part is to describe what isn't working; I explain the options in plain language and in writing.

Is Copilot worth it?

Not always. Before expanding licences I test it with a small group and measure whether it saves time on your actual tasks.

Does our data leave the EU?

It doesn't have to. If processing in the EU is a requirement, I choose a setup that allows it and write into the proposal where data is processed.

Do you work in Estonian or English?

Both, depending on your team.

What does superintelligence (SI) mean, and does it concern our organisation?

Superintelligence (SI) means artificial intelligence that exceeds humans in most intellectual tasks. No such system exists yet, but progress is fast. The value today comes from using AI well in everyday work: tidy data, clear processes and accountability. These are also the preconditions for whatever comes next.

How quickly do you reply?

I usually reply within a couple of working days.

How does an engagement start?

Use the form or write to hello@pearlsroses.com. In the first conversation I clarify your goals and propose a next step, such as an assessment, a solution choice or an integration.

Book a call

In a half-hour conversation we look at where AI would help your work, and agree the next step.

What happens after you write

1

You describe the situation

A few sentences about the process and the goal are enough to start.
2

We talk for 30 minutes

I clarify the goal, the data and the constraints, and where AI will help most.
3

You get a concrete next step

In writing, even if it turns out AI is not the right answer.