01
Find the right use case.
Map the current process, business model, decisions, constraints and data before choosing a model or platform.
- AI opportunity discovery
- Workflow and service design
- Pilot definition
01 / AI consultant
A practical partner from the first question to a working system.
I help businesses identify where AI can genuinely improve the work, then design and build the application, agent or automation around that job.
The same person stays close to the business goal, the interface and the implementation. That keeps decisions clear and prevents a useful idea from disappearing between a strategy deck and a working product.
- 5
- selected builds
- 4
- stages from idea to proof
- 3
- AI service tracks
- 2
- working languages
- 1
- accountable partner
02 / AI consulting and implementation
From a sharp question to a system people can use.
Start with one valuable workflow. Expand only when measured value supports it.
02
Build the right solution.
Create a custom AI application, copilot, AI agent or automation that fits the work around it.
- Custom AI applications
- AI agents and integrations
- Human-in-the-loop interfaces
03
Put it into dependable use.
Turn a compelling demo into an observable system with clear boundaries, responsibilities and a path to improve.
- Testing and evaluation
- Runtime and deployment
- Documentation and handoff
Useful starting points
- Document-heavy work
- Internal knowledge
- Decision support
- Workflow automation
- Focused SaaS products
03 / Selected systems
A portfolio of systems built in practice.
These high-level examples describe the systems and tools without publishing customer data, credentials or private product details.
05selected systems
04delivery layers
00analytics or ad cookies
Local-first AI workspace
Agent OS / Olympus
A workspace connecting agents, tools, memory and production artifacts into one inspectable operating environment.
- TypeLocal-first workspace
- JobCoordinate AI work
- SurfaceAgents, tools, memory
- CoreNext.js + Node
- InterfaceCanvas + console
Privacy-first operations
Iisi platform
A workspace for document-heavy financial operations and review.
- TypeOperations workspace
- JobReview document work
- FocusFinance + documents
- CoreNext.js + Supabase
- TestingPlaywright
Open-source project review
Auto Company
An architecture review of an open-source multi-agent project.
- TypeOpen-source review
- JobAssess multi-agent design
- SourcePublic repository
- MethodArchitecture analysis
- OutputReview notes
Agent operations integration
Hermes operations
An integration reference connecting agent tooling with desktop operations.
- TypeOperations integration
- JobConnect agent operations
- SurfaceTools + desktop
- FocusTool orchestration
- InterfaceDesktop control
Cinematic portfolio system
buildwithllm.dev
This privacy-first portfolio and its cinematic interactive story system.
- TypeInteractive portfolio
- JobExplain AI consulting
- PrivacyNo visitor tracking
- CoreHTML + CSS + JS
- DeliveryCaddy + HTTPS
04 / Tools and languages
Use the tool the system needs.
A practical implementation spans the interface, intelligence, data, runtime and testable proof — not just the model call.
Build the surface.
- HTML
- CSS
- JavaScript
- TypeScript
- React
- Next.js
- Vite
- Tailwind
- Framer Motion
- Three.js
- React Flow
Connect intelligence.
- OpenAI
- Codex
- Claude Code
- Gemini
- MCP
- Ollama
- Hermes
- OpenAI-compatible APIs
Make it operate.
- Node.js
- Python
- Express
- Socket.IO
- Prisma
- SQLite
- Postgres
- Supabase
- RLS
- Docker
- Linux
- WSL
Make the claim testable.
- Graphify
- Playwright
- Vitest
- ESLint
- GitHub Actions
- FFmpeg
- Caddy
- HTTPS
- Oracle Cloud
05 / Working principles
Clarity before complexity.
- 01
Frame the real job.
Define the people, constraints and success measure before choosing an architecture.
- 02
Proof before claims.
Treat tests, runtime behaviour and current limitations as part of the product.
- 03
People remain in control.
Keep consequential decisions understandable and provide a clear review and stop path.
06 / Frequently asked
AI consulting in practice.
A useful engagement starts with the work of the business, not a pre-selected model.
What does AI consulting include?
The work starts with a business goal and the current workflow. We then define a valuable use case, design the solution, build the required AI application, agent or automation, and test it against the real work.
When should a business build a custom AI solution?
A custom solution makes sense when a general tool does not fit an important workflow, data boundary, permission model or human decision point. First, we verify that building creates more value than adopting an existing service.
Can AI consulting start with a small pilot?
Yes. A useful pilot limits the work to one user group, one workflow and a clear success criterion. That makes it possible to assess value, risk and the next decision before a larger investment.
How are privacy and human control handled?
The solution is designed to use only the data it needs. Permissions, logging, model and provider roles, and human review or stop points are defined according to the purpose and risk of the workflow.
07 / Have a difficult workflow in mind?
Bring the hard part.
One messy process, an early AI idea or an application that needs to move from demo to dependable use is enough to start.
Email about a project elias@hakamies.comEmail opens only when selected. This site uses no analytics, advertising cookies, newsletter or visitor profiling.