Attestia - Consulting · AI Advisory

From an AI idea to measurable results. With governance from day one.

I help companies become AI-native. I select use cases with real returns, choose the technology, build AI agents and set out how people should steer them, so they can scale safely, including in banking, insurance and energy.

Book a 30-minute call See packages

Already running an AI agent? Let's check if it's ready →

Looking for an AI governance and AI Act tool only? Go to the Attestia platform →

20+ years in Data & AI · ex-CTO, head of data teams in banking, board advisor · co-creator of Attestia and TrimAlert, creator of Paragrafo, Privuna and TrimAlert - SalesBrain

AI is in every company. Financial impact from it is still rare.

80% vs 37%

respondents in McKinsey's survey who say AI has improved their own productivity vs those who say AI has contributed to their organization's EBIT.¹

42% vs 5%

companies expecting autonomous AI agents by 2030 vs companies with controls in place.²

100%

share of AI leaders in BCG's study ("future-built" companies) with clear ownership of the AI agenda.³

The problem is rarely the AI model. What's missing is process redesign, measurement, ownership, shared data definitions, and rules that say when a person takes over a decision. And an agent only saves time once you subtract the time people spend checking its work. That's what I work on.

One process. One metric. An agent your people know how to steer.

Method: I work with the 6S framework from the book "The Human-Agent Orchestrator" (Bornet et al., 2026). It covers six areas: purpose and the human contribution, business outcome, accountability and constraints, the split of decisions between people and the agent, handover rules, and measurement and improvement. On top of it I use my own workshop tools and an agent value model that counts savings after the cost of oversight.⁴

1
Diagnostic (2–4 weeks)
I find where AI will pay off and what blocks scaling, including whether your data and definitions are ready for agents. You get priorities with a net business case and a 90-day plan. Every use case gets an agent card from the start: purpose and exclusions, owner, metric and autonomy limits.
2
Sprint (6–10 weeks)
I redesign one process and build a working agent on your stack. From its first version the agent records its decisions. Before it moves from the parallel run to the controlled launch, it passes a readiness check: hard scenarios and a stop drill. We measure results before and after. When the agent works on data from several systems, I build its context layer in parallel, as a separate module.
3
Governance (3–6 weeks)
an agent registry with cards, owners, autonomy limits, handover triggers and a decision log. Then regular reviews: is the agent still doing what it was meant to, and has it earned more autonomy?
4
Leadership (3+ months)
as a fractional CTO or CAIO, I keep an eye on the roadmap, costs and vendors. Every quarter we decide which agents to scale, which to fix and which to switch off.

Fixed-price packages

AI-Native Leadership Workshop

a shared view of AI opportunities, a decision on what is worth handing to agents, an agent demo, 3 priorities

For whom: leadership teams of SMEs and corporate divisions
Duration: 1 day
Price: PLN 8,400

AI-Native Diagnostic

7-layer scorecard with a context readiness rating, use cases with an agent card and a net business case, 90-day roadmap

For whom: companies preparing to scale AI
Duration: 2–4 weeks
Price: from PLN 19,600

AI Workflow Sprint

from a process map to a working agent, with a readiness check before launch and measured impact

For whom: companies with a chosen process
Duration: 6–10 weeks
Price: from PLN 50,400with an optional success fee

AI & Agent Governance

agent registry, AI policy, autonomy matrix, handover triggers and the right to stop, initial AI Act classification; optional ongoing agent reviews

For whom: companies deploying agents, financial sector
Duration: 3–6 weeks
Price: from PLN 22,400

Agent Review (Pre-Flight)

already running an agent, even one built by someone else? I check whether it is ready to work under human control and whether it uses your data definitions, and give you a plan of fixes

For whom: companies with a live agent or pilot
Duration: 1–2 weeks
Price: from PLN 8,400

Context Layer for AI Agents

your agent knows your tables but not your definitions? In one domain I work with your data owners to set out terms, metrics and access rules, connect them to the agent and measure its accuracy before and after

For whom: companies whose agents work on data from several systems
Duration: 4–6 weeks
Price: from PLN 33,600context map alone (1–2 weeks) from PLN 11,200

Fractional CTO / CAIO

an owner of your AI and technology agenda without a full-time hire

For whom: SMEs, startups, divisions
Duration: 3+ months
Price: from PLN 6,000/monthpart-time or full-time engagement priced individually

Bank-Ready

preparing a fintech or AI vendor for bank due diligence

For whom: fintechs, RegTech, AI vendors
Duration: 3–5 weeks
Price: from PLN 22,400

Net prices in PLN. Infrastructure, licences and tokens are billed separately and transparently. Every package is also available without my own products.

I usually work as a consultant, but I can also take on a part-time leadership role on your side, for example as a fractional CTO or CAIO.

Who I work with

Companies that already run their first agents or pilots, especially in regulated industries: when agents multiply faster than the rules on who owns them and when a person should take over a decision.
Large companies in transformation: when you'd rather work with an independent expert than a big consulting firm. You need a second opinion, a fast deployment in one domain, or someone who protects your interests with AI vendors.
Mid-sized and smaller companies: when AI assistant licences haven't turned into results and your own team has no time to redesign the process. Fixed price, a specific process, results in weeks.
Startups with strong domain experts: when you know your field but lack a CTO. Architecture, MVP, model selection, and preparation for talks with banks or enterprise customers.

Why me rather than a big consulting firm

Governance from day one: I set the rules for an agent together with the first use case. Decisions nobody makes deliberately still get made — in the tools' default settings.
A practitioner, not a theorist: the first working agent is built during the project, not after it.
Seniority without the pyramid: you work directly with me, from business case to code. Behind me are more than 20 years of data projects: data warehouses, MDM, BI and data platforms in banking, insurance and energy.
Oversight that works: a person at the thresholds and decision categories, not at every step. With the right to stop the agent and evidence that the oversight works. This matters in banking, insurance and energy.
Value counted net: I show savings after the time people will spend overseeing the agent and the risk of errors.
Independent technology choices: I match the technology to your problem and your stack, not to one vendor's catalogue. I compare options against the same criteria.
Speed and transparent pricing: packages from 1 day to 10 weeks, prices published upfront.

I don't just advise. I build.

I built all of the products below together with AI agents. They are in development and user testing, and TrimAlert - SalesBrain is our internal tool. I show them as proof of how I work.

TrimAlert: multi-agent AML alert analysis. Agents gather data from registries and media and prepare an assessment, and an analyst always makes the decision — the system never closes or reports anything on its own. Early Access. Co-created with Agnieszka Czaja. trimalert.com
Attestia: the platform whose website you're on. A compliance OS for companies that build and use AI: AI governance and AI Act compliance — an AI system registry, risk classification, documentation and an audit trail. In advisory projects, also an agent registry and agent cards. I built the first version in about 40 days. Co-created with Agnieszka Czaja. Go to the platform
Paragrafo: an iPhone scanner for risky contract clauses. On-device analysis, no cloud AI. App Store
Privuna: a privacy check before you send a document. Sensitive data is detected and redacted on the device. App Store
TrimAlert - SalesBrain: a knowledge and procedure base where our AI agents work every day: approved knowledge with source and date, procedures for agents and automated quality checks. It runs in Claude Code and Codex. It is a context layer at small scale, and the Context Layer for AI Agents module follows the same principles.

Disclosure policy: when I recommend one of these products, I say clearly that I created or co-created it. I always present an alternative, and my advisory fee doesn't depend on your choice.

AI under control: banking, insurance, energy

In regulated industries an AI agent needs an owner, autonomy limits, a decision log, and clear rules on when a person takes over. Having a person in the process is not control in itself: what counts is whether they have the time, the information and the right to stop the agent. I help design and document this:

an inventory of AI systems and agents, and an initial risk classification based on the AI Act categories;
a card for every agent: purpose, owner, constraints, handover triggers, manual fallback and retirement criteria;
an AI vendor risk assessment you can use in your own compliance processes, e.g. for DORA;
documentation and evidence for internal audit: who owns each agent, what it is allowed to do, how it reached a decision, and whether the stop drill worked.

Disclaimer: I don't provide legal advice. Classifications and documentation are expert support; legal interpretation should be confirmed by your legal team or law firm.

Advisory and the Attestia platform:

The Attestia platform: when you want to keep a registry of your AI systems on your own, classify them and prepare AI Act documentation in a few hours. You can start with the free plan. Go to the platform
Advisory: when you need more than documentation: agent oversight, process redesign, building AI agents, or an owner of your AI agenda.
Disclosure: Attestia is my product. In the Governance package we can keep the agent registry and agent cards in Attestia or in your own tools, and the package price is the same. Pre-Flight checks, agent reviews and the quarterly board report can run in Attestia too. At the end of the project I hand the Attestia organisation over to you or export everything to DOCX, XLSX and PDF. You can prepare AI Act classification and documentation in Attestia, in another tool, or on your own. I compare these options against the same criteria, and my fee doesn't depend on your choice.

Grzegorz Czaja

For more than 20 years I've connected business with data and technology. I was CTO at Algomine, head of Banking Data teams at Santander CIB and a Data & AI advisor at Predica. For six years I ran data warehouse, BI and ML programmes for PGE Group. Most recently, as Technology Advisory Manager at Elitmind, I advised executive boards on Data & AI strategy and sat on steering committees. I started as a software developer and data engineer at Citi and AXA.

Today I run AI advisory as Attestia - Consulting and, together with Agnieszka Czaja, develop the TrimAlert and Attestia products. I hold master's degrees in management and computer science from the University of Warsaw.

What former managers and colleagues say about working with me: recommendations on LinkedIn →

FAQ

How are you different from a big consulting firm?

You work directly with a senior expert who builds products and agents himself. We start with one process, and prices are fixed and known upfront. Oversight rules are set from day one, not added as a separate project at the end.

We already run an agent. Where do we start?

With an agent review. In 1–2 weeks I check whether it's clear who owns the agent, when a person takes over a decision and how to stop it. I also test whether the agent uses your data definitions. You get a list of gaps and a plan of fixes, also for an agent built by someone else.

What is the 6S framework?

A framework from the book "The Human-Agent Orchestrator" (Bornet et al., 2026). It covers six areas: purpose and the human contribution, business outcome, accountability and constraints, the split of decisions, handover rules, and measurement and improvement. Based on it, I prepared my own workshop tools and an agent value model.

What is a context layer for AI agents?

The definitions of terms and metrics, data owners and lineage, access rules and procedures that an agent uses while it works. Without it, an agent knows the structure of your data but not its meaning, so its answers are technically correct and wrong for the business. I start with one domain and with the agent's decisions, not with a catalogue of every table.

Which tools do you use for the context layer?

First I check what you already have, e.g. Fabric IQ, Unity Catalog or Power BI semantic models. If you need a new tool, I compare options against the same criteria, including open source such as OpenMetadata. A company-wide data catalogue rollout is run by your team or a partner.

Do I have to use the Attestia platform?

No. Attestia is my product and I recommend it only when it fits. I always present an alternative, and every package costs the same with or without Attestia. If we work in Attestia, after the project you keep read access to the agent cards and decisions, plus a full data export.

Do you advise or implement?

Both. In the Sprint I build a working agent on your stack. Large-scale rollouts and ongoing maintenance go to your team or a partner.

How do you charge?

A fixed price per package, a monthly retainer for the CTO/CAIO role, and an optional success fee in the Sprint. The fee is based on the business result together with the quality of the agent's work, and is settled after measurement in the first weeks of operation. Infrastructure and token costs are separate and transparent.

Do you recommend your own products?

Only when they fit, always disclosing that I created or co-created them, and always alongside an alternative. You can order any package without my products.

Do you work with small companies?

Yes. The smallest packages are a one-day leadership workshop and an agent review.

Can you be our CTO or CAIO?

Yes. Usually on a fractional basis, a few days a month. A part-time or full-time commitment is also possible.

How do you handle confidentiality and data?

I work under an NDA. We agree on data and system access at the start, and I build agents on your infrastructure.

What is the first step?

A free 30-minute call. After that I propose a package and scope.

Let's talk about your first process

Describe in 2–3 sentences what you want to achieve with AI.

gczaja@trimalert.com · LinkedIn

Attestia - Consulting · Grzegorz Czaja

Sources:

McKinsey, The state of AI in 2026, pp. 11–12;
BCG, The Formula for Agentic AI Value, 2026, p. 15;
BCG, The Formula for Agentic AI Value, 2026, p. 12;
P. Bornet, J. Wirtz, T. Stephens et al., The Human-Agent Orchestrator: Leading and Scaling AI-Driven Organizations, 2026. The workshop tools and the agent value model are my own work, not material by the authors.

Studies 1–3 are global, not specific to Poland.