We design the decision processes inside operational teams that price, source, plan capacity, and allocate capital. The AI systems we build let those teams act on the full spread of outcomes.
We treat AI as a booster for well-designed operations, not a shortcut around unclear data, weak models, or missing structure. We measure success in variance removed from operational decisions, hours recaptured, and bids that price exposure correctly.1
§ Methods · how we think about a decision
Fig. I
Probability of outcomes
Size decisions to the distribution.Fig. II
Decisions, branched
Make the choice tree explicit, even when it's small.Fig. III
Value of information
Cheap signals first. Expensive ones, only if they pay.Fig. IV
What moves the answer
Spend modeling effort where the bar is widest.
Operational surfaces.
We work where decisions are repeated, instrumented, and consequential. The boring loops that compound. The list below is illustrative. The full list is the conversation we have with you.
Supplier selection under uncertainty, with total-cost models that include lead-time risk and quality variance.
P10–P90
Forecasting
Demand and capacity as full distributions. Plans sized to the spread, with a deliberate worst-case.
h/wk
Supply chain
Routing and prioritization under capacity constraints. The boring loops where most operating cost lives.
min/case
Customer service
Triage, prioritization, exception handling. Faster routing of the cases that actually need a human.
σ planning
Finance
Forecast review, scenario sizing, capital allocation. Distribution-aware planning with explicit trade-offs.
§ Practices
AI capability uplift
We train your teams to build custom AI workflows: identifying value-adders, surfacing relevant information, integrating tools deeply into existing software, and governing autonomous agents.
Engagement structure: 12 weeks. Ends with a working system and a team that owns it.
Operational software
[ DEV ]
We are building software for specific industries and operations, designed to plug into existing tools and data and solve a clear, measurable problem very well. First releases targeted for late 2026.
Engagement structure: pilot partnerships, 2 to 4 design partners per product.
§ About
Who we are.
Selkio is a small partnership based in Helsinki. Our backgrounds are in decision analysis, operations research, and AI, and in running the kinds of operating teams we now build for.
Our work focuses on integrating AI into the processes where it materially shifts outcomes. We also help clients apply modern mathematical modelling to sharpen strategy and operations.
