Applications overview
Know which AI pilot to trust on Monday
Seven application types, each lined up against the Estonian example that shows it working — and the one thing worth checking before anyone in your organisation relies on it.
- Guidance that ends in a next step
- Training built from your own manuals
- From a sentence to an app draft
- Optimisation on the factory floor
- Help inside official channels
Five rows shown — read the full registry below
The registry
Which AI application solves your task?
Practical AI application types, grouped by the problem each one solves, with a working Estonian example beside each group. You will see career guidance, internal training, app creation, industrial optimisation, sports coaching, biosensing research, and citizen assistance. Every entry stays inside what has been reported; where something is still development work, this page says so.
- Career guidance moves from search to action Roleo · Tallinn
- Training that reads your own manuals PocoBit · Estonia
- From a sentence to a working app draft Bilt · prompt to product
- Boring optimisation with visible value Grainflow · agriculture
- Coaching built on training data KULG · endurance sport
- Biosensing at the edge of health data Muun Health · Tallinn
- Citizen help inside government channels Bureaukratt · RIA
Guidance
Career guidance moves from search to action
Roleo, based in Tallinn, is developing an AI career agent. It was reported to have raised 2.9M euros and aims to expand across Europe. The angle here is the move from job search to active guidance: the agent helps structure options and next steps rather than handing over a list of postings and leaving the work to the reader.
For jobseekers and HR teams the test is simple. Does the guidance stay tied to the skills the person actually stated and the roles genuinely open in their area, or does it drift into generic advice anyone could give? Keep a person in the loop for the final decision — an agent that helps prepare a case is useful, an agent that decides on someone's behalf is a liability.
- Good fit
- Candidates who want structured preparation and follow-through across several weeks, not a one-off search.
- Check before you rely on it
- Which sources feed the role suggestions, how candidate profiles are stored, and who reviews an output before it reaches the candidate.
A working career session in Tallinn: the agent supports the conversation, the counselor still leads it.
What the four rows above have in common
Every one of them replaces a habit, not a person
Career guidance replaces a browse. Training replaces a static PDF. Prompt-to-app replaces a first scoping meeting. Dryer optimisation replaces a manual set-point. Reading them together, the pattern is the same: the tool takes on the repeated, tedious step, and a person keeps the judgement. That is what makes a pilot survivable in an Estonian organisation — it lands in a job someone already does, on a schedule that already exists.
Onboarding across shifts and sites — the training module follows the manual, so the manual has to follow the work.
Training and building
Training that reads your own manuals
PocoBit turns company knowledge and internal documents into interactive training experiences. It targets knowledge retention and aims to streamline employee training, and it is built for organisational learning rather than public content. For teams in Estonia running onboarding across shifts or sites, that replaces a static PDF with practice and checks.
The upkeep rule decides whether it survives a year. When a procedure changes, the training source has to change the same week — otherwise the organisation trains people on the version it no longer uses. Ask who owns the source files and how corrections actually reach the learner before you commit a team to it.
- Good fit
- Operations teams with repeat procedures and a rotating workforce: warehouse, plant, service desk, regulated care work.
- Check before you rely on it
- Who updates the source files, how a corrected procedure reaches the next cohort, and what happens to a learner mid-course.
Building
From a sentence to a working app draft
Bilt turns a text prompt into a native mobile app and ships apps to the major mobile app stores. It represents prompt-to-product automation for fast app creation by people who are not full-time developers. Treat it as a way to test an idea with a usable draft, not as a replacement for review, testing and store policies — the store rules do not relax because the code arrived quickly.
Teams still have to confirm permissions, data handling and maintenance ownership before release. Whoever is on the hook for the next security update should be a named person before the app leaves the draft stage.
- Good fit Small firms testing a service companion app or an internal tool without a full development team.
- Check Store requirements, data flows, and who maintains updates after the second release.
Industry
Boring optimisation with visible value
Grainflow develops AI-powered technology for grain drying. Its guidance solution optimises dryer performance and aims to reduce costs and improve profitability for agricultural businesses. This is AI in heavy industry and farming operations, where a small efficiency gain compounds across a whole season. An Estonian farm runs dozens of drying cycles; a few percentage points per cycle add up to real fuel.
The lesson for other sectors: look for processes with steady sensor data and clear cost drivers. Those show results sooner than open-ended chat pilots, because there is a measurable thing to move and a number to move it against.
- Good fit Agricultural and process operations with continuous instrumentation and a clear per-unit cost to reduce.
- Check Which sensors are already reliable, and whether the operator can still override the system's own recommendation.
Send a request
Ask about one application type
Tell us which row is closest to your task and we will come back with what we would check first, what the honest effort looks like, and which Estonian example is nearest to your situation.
The five checks we walk for any application type
- Who is accountable for the output once it leaves the system?
- Which source data feeds it, and who keeps that data current?
- Where does a person still sign off before the output is used?
- How does the system behave the week after a process changes?
- What is the exit — what happens if the pilot does not continue?
The same five questions sit behind our assessment method. If any of them is still open, the pilot is not ready for Monday — however impressive the demo was.
See how we assess solutionsEndurance sport
Coaching built on training data
KULG develops a training platform for runners and coaches. It uses training data to provide adaptive guidance, aiming to support sustainable progress and reduce injury risk. It was noted as part of a 2026 AI growth program with a sports-tech angle. Unlike generic wellness tracking, the emphasis is on coaching interaction over time — Runners and coaches in Estonia can read this as an example of domain-specific guidance done carefully.
Health data
Biosensing at the edge of health data
Muun Health was founded in Tallinn in 2023 and is developing wearable biosensors for continuous hormone monitoring. It sits in Estonia's health-data and biosensing niche. This is development work, not a finished consumer outcome — readers interested in health tech should follow verified company updates for status rather than treat an in-development product as a decision input today.
Continuous biosensing research in Tallinn: an in-development product, not something to rely on this year.
Public sector
Citizen help inside government channels
Bureaukratt was developed by Estonia's Information System Authority (RIA) and is positioned as a virtual assistant for citizens. It forms part of Estonia's broader public-sector AI push. The idea is help inside official services, with attention to secure handling — you ask a question in the channel you already use, and the assistant points you toward the correct procedure.
Reason Reserve is listed by RIA as an innovation project for secure AI use in the public sector. It focuses on enabling AI assistants across Estonia's government systems and belongs to infrastructure-level strategy. It is a controlled environment rather than a consumer app. Pair the two in your notes: Bureaukratt is the visible assistant; Reason Reserve is the governance thinking behind safe deployment. Our public-sector page adds the procurement context.
The board's own working set — what each closed cycle produced
01
A named next step for a jobseeker
02
A training module kept in step with the manual
03
A draft app with a named maintenance owner
04
A measured drying cycle with an operator override
05
A public-sector assistant inside its official channel
Where to go next with your own task
If your task runs through an internal process — quoting, invoicing, dispatch, scheduling — the business automation page carries the operational detail. If it touches a public body, the public-sector page covers procurement and secure handling. And if you want the criteria before you talk to anyone, the assessment method is published in full.
BOLTCAT OÜ · Kastani tee 6-1, Laitse küla, Saue vald, Harju maakond, 76302, Estonia · Monday to Friday, 9:00–18:00.