{"id":1202,"date":"2026-08-19T08:50:00","date_gmt":"2026-08-19T08:50:00","guid":{"rendered":"https:\/\/nexpert.fi\/?p=1202"},"modified":"2026-08-19T10:59:59","modified_gmt":"2026-08-19T10:59:59","slug":"ai-piloteista-kayttoonottoon","status":"publish","type":"post","link":"https:\/\/nexpert.fi\/en\/ai-piloteista-kayttoonottoon\/","title":{"rendered":"From AI pilots to production"},"content":{"rendered":"

Tips on how to get to the next phase in use<\/h2>\n\n\n\n

In my previous post, I highlighted Master's theses as one good way to conduct various surveys, experiments, and development projects within organizations. You don't always need heavy consultant procurement or a months-long project. A well-defined thesis can produce both a description of the current state and concrete proposals to move forward surprisingly quickly. But what if the organization wants to undertake the mapping process on its own?<\/p>\n\n\n\n

One current topic is the next phase of artificial intelligence usage. In the 2025 Work Trend Index, Microsoft described organizations' journey toward the so-called Frontier Firm model through three phases. In the first phase, AI acts as a human assistant. The human still does the work, but AI helps do the same work faster or better. In the second phase, agents join teams as digital colleagues and take responsibility for entire tasks under human guidance. In the third phase, agents can already handle entire processes while the human sets the direction and intervenes when necessary.<\/p>\n\n\n\n

Many organizations are currently somewhere within the first phase. You ask ChatGPT a question. You write an email with Copilot. You use AI to summarize meeting notes or ask for help with an Excel formula. All of these are completely sensible use cases. But there is still quite a long way from them to AI genuinely changing the way work is done.<\/p>\n\n\n\n

But how do you know what you should do next?<\/h2>\n\n\n\n

In the 2026 Work Trend Index, Microsoft published what I think is a quite useful four-quadrant matrix for this.<\/p>\n\n\n\n

\"Nelj\u00e4<\/figure>\n\n\n\n

In the four-quadrant model, the use of artificial intelligence is divided into four ways: Asking, Experimenting, Collaborating, and Delegating. The model examines both how actively humans participate in the work and how much the AI or agent does the work in the background.<\/p>\n\n\n\n

Asking is already very familiar to us. We ask about something, ask to format a paragraph better, or change data in a table into another format. In collaboration with artificial intelligence, we are already truly working together. For example, an analysis is created through several rounds. The AI makes a suggestion, the human evaluates it, provides more information, and the AI continues the work.<\/p>\n\n\n\n

The experiment tests what AI is capable of. Could it, for example, handle a workflow that is new to us? Can the agent retrieve the necessary information from various sources? At what point does it need human help? Delegation is perhaps the most interesting aspect when thinking about the transition toward phase 2. In it, the human defines the goal, but the AI does the actual work. For example, it can compile a recurring report from various inputs, format the data into an agreed-upon structure, and bring the final result to the human for review. At this point, we are already talking about more than just better prompts.<\/p>\n\n\n\n

Try this in your own organization<\/h2>\n\n\n\n

I wouldn't start by thinking about what agents should be built for the company, but rather figure out what people are already doing with AI.<\/p>\n\n\n\n

A very light mapping could go like this:<\/p>\n\n\n\n

    \n
  1. Ask a few employees to log concrete situations over a week or two where they use artificial intelligence. Not at a general level like \u201dI use Copilot,\u201d but what was actually done with it.<\/li>\n\n\n\n
  2. Map the use cases into Microsoft's quadrant. Was it a quick question, collaborating, experimenting, or delegating work to AI?<\/li>\n\n\n\n
  3. Look particularly at repetitive tasks. What is done weekly or monthly over again? Where is information gathered from multiple sources, copied from one system to another, summarized, or prepared for decisions?<\/li>\n\n\n\n
  4. Select one or two use cases where work could be shifted from the Asking or Collaboration box towards Delegation. It is not advisable to start developing multiple use cases at once.<\/li>\n\n\n\n
  5. Only investigate after that what the implementation requires. Where does the data come from, which systems the AI needs to access, what it is allowed to do independently, at which point a human reviews the work, and who is responsible for the final result.<\/li>\n<\/ol>\n\n\n\n

    At this point, you usually start finding the first genuinely sensible agent use cases. For example, creating a monthly report might currently mean that a person fetches figures from three systems, transfers them into Excel, performs a few calculations, writes a summary, and sends the report forward. In phase 1, AI might help write the summary. In phase 2, we should ask whether the system could fetch the data, perform the initial analysis, identify anomalies, generate the report, and present it to the human for review. The difference between these is significant, and the potential for saving time is large.<\/p>\n\n\n\n

    The four-quadrant model is not a race toward the most autonomous AI possible<\/h2>\n\n\n\n

    This matrix is not a maturity model for ways of working where all tasks should eventually end up in the delegation box. What distinguishes advanced AI users is rather their ability to assess which way of working suits each task. Not everything is worth automating.<\/p>\n\n\n\n

    If a task requires a lot of human judgment, discussion, trust-building, or situational assessment, collaboration may be just the right way to use artificial intelligence. But if the same person spends two hours every Friday retrieving the same information from the same places and compiling it into the same report, it is worth pausing. Could that be done differently?<\/p>\n\n\n\n

    Understanding the current state before technology<\/h2>\n\n\n\n

    Here we actually arrive at the same thing as in many other development projects. First, you have to understand what is currently being done. Only after that is it worth deciding what to change and with what technology.<\/p>\n\n\n\n

    The mapping can easily be done yourself within your own organization. It can be done as a workshop. Alternatively, it can be structured as, for example, a Master's thesis or another RDI project, if you want to investigate the current state a bit more thoroughly and at the same time obtain a documented development proposal.<\/p>\n\n\n\n

    The most important thing is to move beyond the question, \u201cWhat are all the ways we could use AI?\u201d<\/strong> and start asking, \u201c\u201cHow is this work done today, and which part of it would make sense to have AI handle?\u201d <\/strong>
    That\u2019s where the journey from Phase 1 to Phase 2 actually begins.<\/p>\n\n\n\n

    Would you like to assess where you stand?<\/h2>\n\n\n\n

    If you want to find out how AI is currently being used at your organization and where the next logical steps toward phase 2 can be found, we can carry out an assessment together. We will review current use cases using Microsoft's matrix and identify the work phases where AI could already do more than just assist. Together, we will look at where you should get started.<\/p>\n\n\n\n

    \n
    Book a free consultation meeting<\/a><\/div>\n<\/div>","protected":false},"excerpt":{"rendered":"

    Vinkit miten k\u00e4yt\u00f6ss\u00e4 p\u00e4\u00e4see seuraavaan vaiheeseen Edellisess\u00e4 kirjoituksessa nostin esiin YAMK-opinn\u00e4ytety\u00f6t yhten\u00e4 hyv\u00e4n\u00e4 tapana tehd\u00e4 organisaatioissa erilaisia selvityksi\u00e4, kokeiluja ja kehitt\u00e4mishankkeita. Aina ei tarvita raskasta konsulttihankintaa tai kuukausien projektia. Hyvin rajattu opinn\u00e4ytety\u00f6 voi tuottaa yll\u00e4tt\u00e4v\u00e4n nopeasti sek\u00e4 nykytilan kuvauksen ett\u00e4 konkreettisia ehdotuksia eteenp\u00e4in. Mutta ent\u00e4 jos organisaatio haluaa l\u00e4hte\u00e4 tekem\u00e4\u00e4n kartoitusta itse? Yksi ajankohtainen aihe […]<\/p>\n","protected":false},"author":6,"featured_media":1296,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[31],"tags":[213,218,212,217,216,215,42,219,214,136],"class_list":["post-1202","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-prosessit-kuntoon","tag-ai-pilotti","tag-delegointi","tag-kayttoonotto","tag-kokeilu","tag-kysyminen","tag-nelikentta","tag-tekoaly","tag-tyonkulku","tag-work-trend-index","tag-yhteistyo"],"_links":{"self":[{"href":"https:\/\/nexpert.fi\/en\/wp-json\/wp\/v2\/posts\/1202","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/nexpert.fi\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/nexpert.fi\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/nexpert.fi\/en\/wp-json\/wp\/v2\/users\/6"}],"replies":[{"embeddable":true,"href":"https:\/\/nexpert.fi\/en\/wp-json\/wp\/v2\/comments?post=1202"}],"version-history":[{"count":6,"href":"https:\/\/nexpert.fi\/en\/wp-json\/wp\/v2\/posts\/1202\/revisions"}],"predecessor-version":[{"id":1298,"href":"https:\/\/nexpert.fi\/en\/wp-json\/wp\/v2\/posts\/1202\/revisions\/1298"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/nexpert.fi\/en\/wp-json\/wp\/v2\/media\/1296"}],"wp:attachment":[{"href":"https:\/\/nexpert.fi\/en\/wp-json\/wp\/v2\/media?parent=1202"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/nexpert.fi\/en\/wp-json\/wp\/v2\/categories?post=1202"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/nexpert.fi\/en\/wp-json\/wp\/v2\/tags?post=1202"}],"curies":[{"name":"WP","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}