Tulevaisuuswiki / Työn tulevaisuus
AI Work Matching
Also known as Personal Labour Market, Tekoälyvälitys
AI Work Matching is the continuous allocation system used in the Jobi Economy to match people's needs with other people's abilities. It identifies unmet needs, models each participant's skills and preferences, and offers participants ranked menus of tasks. Its functions were first developed separately in the many AI-assisted task networks that appeared from 2028; they were federated into a common national matching layer in 2031–2032. The system replaced conventional job search for most of the Finnish working-age population during the early 2030s.
Background
In conventional labour markets workers search for positions and employers search for workers. Economists describe this process as costly matching, and the costs are tolerable when work is bundled into long-term positions. For millions of small tasks they are prohibitive: searching for, vetting and contracting a suitable person often costs more than the task is worth.
AI systems reduced these search and transaction costs to near zero, making even very small exchanges economically viable. Analysts have compared the effect to that of earlier peer-to-peer platforms, which allowed small suppliers to participate in markets previously dominated by firms, but at a considerably finer scale.
Development
In the fragmented early networks, each service used its own matching logic, profiles and ranking criteria, typically optimizing for the number of completed tasks or the network's fee revenue. Participants' personal assistants, which searched several networks at once, formed the first cross-network matching layer. The Act on Jobi Transactions (2030) required portable participant profiles and a common exchange protocol, and between 2031 and 2032 the separate matching services were federated into a single national matching layer. The Act on Algorithmic Work Allocation (2032) replaced the networks' differing objectives with a common one.
Operation
Need discovery
The system does not rely solely on participants posting tasks. With the participant's consent, personal AI assistants identify probable unmet needs from calendars, conversations, maintenance records, shopping lists and past behaviour. A participant who has never announced a willingness to pay someone to organize their garage may be asked whether they would do so for two jobis; if they confirm, a new transaction becomes available immediately.
Participant model
For each participant the system maintains a continuously updated model covering:
- demonstrated skills and credentials;
- physical capabilities;
- reliability;
- location and available equipment;
- learning history;
- stated and observed preferences;
- willingness to perform different kinds of work.
The model functions as an estimate of the participant's comparative advantage.
Ranking
Participants receive ranked menus of tasks rather than a single assignment. Ranking takes into account:
- compensation;
- predicted enjoyment;
- probability of successful completion;
- travel and preparation costs;
- long-term value of the skills involved;
- effect on future earning opportunities.
Because of the last two factors, a recommendation may be suboptimal in the short term but optimal over the participant's working life (see Lifetime Earnings + Fun).
Difficulty management
The system deliberately offers work near the limit of a participant's existing competence. Simple tasks establish reliability, and more demanding tasks introduce adjacent skills. Where appropriate, the AI provides real-time instruction during the task. Researchers have described the effect as turning much of the labour market into a continuous training environment (see Paid Learning).
Choice
Participation is voluntary at the level of individual tasks: the system recommends and the participant decides. A typical menu includes several categories of offer:
- familiar, low-risk work;
- higher-paid work;
- work predicted to be enjoyable;
- developmental work involving new skills;
- highly paid work that the system predicts the participant will dislike.
Usage data indicate that the last category is accepted considerably more often than early designers expected.
Regulation
AI Work Matching operates under the Act on Jobi Transactions (2030) and the Act on Algorithmic Work Allocation (2032). The latter codified the allocation objective, required the system to explain its recommendations on request, prohibited the use of protected characteristics in ranking, and established the right to exploration described in Paid Learning. Matching infrastructure is operated by a public-private consortium and funded partly through the Verus.
Criticism
The system has been criticized on grounds of:
- privacy, particularly in connection with need discovery;
- discrimination arising from historical data;
- manipulation of participants' choices through ranking;
- excessive dependence on algorithmic recommendations;
- self-reinforcing skill profiles that limit occupational mobility;
- concentration of economic coordination in a small number of systems.
These concerns shaped the Act on Algorithmic Work Allocation and remain central to Finnish political debate on the Jobi Economy.