HEA Group

We make companies AI-native.
With agents from idea to production.

HEA Group helps companies with architecture, engineering and security as AI agents take on more of the work. We build, bring the way of working into your team and check that it holds up.

Book an intro callSee what we do

How we build

AI writes, tests, reviews and operates. People come in where judgement and accountability are actually needed: what to build, which risks are acceptable and what happens when something is new.

StepHow it worksWhere people come in
IntentAgents gather the context, challenge the requirements and write a spec with goal, acceptance criteria and risks.Set the goal and what counts as done.
BuildSeveral agents build in parallel, in small isolated steps, with tests for every acceptance criterion.Only if the goal changes.
VerificationTests, type checks, security scanning and review by several models. The agents fix and rerun until everything passes.Set the bar once. Nobody reads every line.
Review by riskRoutine changes reach production on evidence. Changes to architecture, dependencies, security or personal data are raised as a targeted question.Answer the question, not the whole pull request.
OperationsAgents watch logs, traces and security signals, tie anomalies to the change behind them, and roll back or isolate.For interventions beyond the playbook.
LearningEvery finding, incident and decision becomes a rule, a test or an architecture decision record, so the same question never needs asking twice.Not at all. It happens automatically.

Services

Architecture and technical advisory

Senior architecture and CTO advisory for companies building or renewing their systems. We do not lock you in to one vendor: we evaluate models and platforms against your own tasks, choose by what your data, requirements and budget call for, and build so you can switch when something better arrives.

  • Architecture and technology choices for new and existing systems
  • AI models from several vendors, open-weight included, evaluated against your tasks
  • Cloud or on-prem, EU or US cloud, depending on what your data and regulation require
  • CTO advisory, information security and AI governance

AI-native engineering

We build products and internal tools with teams of AI agents, backed by automated checks for quality, security and cost.

  • New products from idea to production
  • Agents and workflows in existing systems
  • Infrastructure as code in the cloud or environment that suits you

AI-native operations

We help you move work from manual steps to agents with clear boundaries: engineering, marketing, finance and operations.

  • Mapping where AI helps and where it does not
  • Ways of working, tools and rules for your team
  • A pilot in one area, then the next

Training

Hands-on workshops for engineers and leadership. We work in your own repositories and processes, not in made-up examples.

  • Agent-driven development for engineering teams
  • Opportunities and risks for leadership teams
  • Guidelines for safe use of AI

Security when attackers have AI

Models that find and exploit vulnerabilities are now on both sides, and open-weight models can run with no safeguards at all. Vulnerabilities are often exploited before a patch exists, and findings arrive faster than most teams can fix them. What decides the outcome is how fast you find, fix and detect, across the whole development and operations cycle.

This offering is new. We are looking for our first pilot clients.

Find it before the attacker

Continuous AI-driven pentesting and code analysis on every change and release, not once a year. We use the same class of models attackers use, open-weight ones included.

Fix in hours

Agents write the fix and a regression test, and the attack is rerun to confirm the hole is closed. The fix takes the same risk-routed path to production as all other code. We measure time from finding to fix in production.

Supply chain

Dependencies, container images, build pipeline, AI skills, MCP servers and models. Continuous inventory, signed builds and automatic updates that are verified before they ship.

Detection and response

Logs, traces and security signals in the same view as your releases. Agents triage alerts, tie them to the change that caused them and contain the damage at once.

Identities and keys

Short-lived credentials instead of static keys, least privilege for people and agents alike, and automatic rotation when something leaks.

Secure AI features and agents

Prompt injection, data leakage and over-privileged agents. Isolated runtimes, filtered network traffic and local models when data must not leave the building.

Talk to us about a pilot

Our products

Products we have built and run ourselves, with the same method we use in client work. Nummerlappen is ready to use today, Retro is open to try in beta, and GetAPrompt is in beta by invitation.

About us

HEA Group AB was founded in 2026 in Stockholm by Niklas Hellberg, who has 25 years in architecture and security, including national infrastructure and BankID, the Swedish digital identity.

Today HEA is one founder and a team of AI agents. You get one senior person accountable for the whole, and agents that make it fast. We take on engagements where there is a clear owner on your side and a goal that can be measured.

Niklas on LinkedInNiklas’s blog on AI-driven development (Swedish)

Tell us what you want to achieve

Write a few lines about your situation, and we will get back to you with a proposal for a first conversation. We often start with a quick MVP or pilot in a short engagement. Scope and pricing we settle in that conversation.