STAIR: A research-backed methodology for adopting AI

Full session transcript

THIS IS AN AUTOMATED TRANSCRIPT

Noz: Hello everybody. My name is Noz. I am your host here on the OmnichannelX podcast. I’m very excited about this episode. I have with me today Martin Lassen-Vernal and Louise Harder Fischer. Louise is an associate professor at the IT University in Copenhagen, and Martin is head of communications in the City of Copenhagen administration, and also an external lecturer at the IT University of Copenhagen. We have been chatting on and off for a little over a year now. Today we’re going to talk about some methods that Martin and Louise have been collaborating on, and how they might be applied when looking at our AI projects. I’ll let you two summarise the goals. Louise, would you like to introduce yourself first?

Louise: Yes. I’m an associate professor at the IT University in Copenhagen. Being an associate professor, you also do a lot of research, so you immerse yourself in practice. During the last two to three years, my main research question has been: how do you make work meaningful for people in the AI era? I’ve been working with Martin and his department, collaborating on an action-oriented method to actually meet that goal of creating meaningful work in the AI era. That’s pretty much what I’m doing.

Noz: Great. And Martin.

Martin: I’m head of a communications department. We’re around 20 to 25 people, in one of the administrations of Copenhagen. I’ve actually been working with AI since early on — not long after ChatGPT went live. We embarked on a programme in my administration, and during that process we met Louise, because we were looking for a new methodology. It was obvious to us that we needed something different in terms of how to work with this new technology. So the main focus has been on meaningfulness, the kind of value we produce, and how that guides why and how we use AI.

Noz: So what is this method called, and why did you create it? Louise, do you want to start?

Louise: The method is called STAIR, which is an acronym: S-T-A-I-R, standing for Social Technical AI Reflection. I come from a sociotechnical perspective — that’s the mindset through which I understand phenomena in organisations and the changing nature of work with ICT. So that’s the ST element. Martin can elaborate on the AIR element.

Martin: The sociotechnical methodology could really be applied to any sort of technological phenomena, any kind of digitalisation process. But in the context of AI, and specifically generative AI, there is quite a lot that is different from earlier technologies. So the AI part emphasises that. And Reflection — I think that’s a well-chosen word. It’s the meat of the matter. It’s the actual content of the method. What we put into a structured framework. We call it structured reflection.

Noz: Fantastic. So it’s structured reflection. Great.

Martin: Yes. There is a methodology here, and there are guardrails around it. We invite open-endedness and reflection, but it’s structured around a set of principles and insights from the research and our practical learning programme.

Noz: I think it’s valuable because many of our listeners will be wondering how they, within their own job roles — and if they’re team leaders or managers — how they maintain meaningful work for their staff. How do we stop ourselves from simply becoming, as I’d put it, meat puppets: there to review AI output and make sure the machines keep humming? I’m very interested in process reimagination in this age. There’s an instinct I’m seeing from clients that just says: can’t the AI just do it? Why do we need to do all this thinking, now that we’ve invented a thinking machine? That bothers me. Because the classic approach to technology introduction is often: who can I fire? Companies often choose to do the same thing cheaper, rather than think about how they could do better. I use the phrase “value, not volume” a lot. My listeners might be getting sick of it. But I feel like what you’re talking about gives a bit more method to that. Even if management says they have to cut headcount, somebody’s going to be left. So how do we build a process? What are the components of this methodology? What does the structure look like?

Louise: I’d say first that the method builds on a couple of theoretical perspectives. It combines work design — how organisational designers and workplace psychologists think about job identity and the outcomes of redesigning work processes — with job crafting, which is another layer of understanding how work design evolves, but in the hands of the individual. Job crafting is how individual people redesign their work so it fits them and their preferences better, in order to be more productive and to reach elements of wellbeing. There’s work design from the top down, and job crafting as the bottom up. Both elements come from a sociotechnical perspective where you rely on practice. The sociotechnical mindset is about not just the technology, but having sensitivity around how you design the organisational system around the technology. Then Martin can put that into the more concrete, practical steps.

Martin: Yes, that’s exactly right. We looked at Scrum and the Agile methodology and terminology. What we were looking for was a method that puts these things into practice. I’d say there are macro steps to STAIR. The first is defining your principles — what are the working principles that guide reflection? Principles can be framed as questions, as points you want to observe, as prompts that guide conversation around AI. We can come back to how we actually practise this reflection. But the first step is: what are your principles?

Noz: Can you give me an example?

Martin: A principle could be: when we use generative AI in our work or workflows, we want to maintain value in work. That’s a common principle. Each principle has a set of sub-questions and points of interest that you want to dig into if you’re what we call a STAIR Master — again borrowing terminology from Scrum.

Noz: I’m going to ask a simple question with a complex answer. What’s value?

Martin: Exactly. That’s actually the question the STAIR Master would ask. What kind of value does this produce? Is it in terms of economics, efficiencies, quality?

Louise: Professionalism. Does it enhance your professional identity?

Martin: So there are many ways you could talk about this. We don’t use these principles as a predefined set of rules, but we want to be aware of what kind of value proposition we have. We can come back to how we’ve actually shifted the value proposition during these conversations. Other examples: do we have frameworks and guidelines? Are we addressing compliance, quality, experimentation, learning? You can structure this as you want. In our case, we have eight principles. But again, the first step is to define your own principles. As Louise would say, the insights from sociotechnical theory would inform some of those principles. But you’d also have your own context, culture, and language that you want reflected.

Noz: Let’s go through what the framework looks like, and then maybe circle back on how it can be adapted. You said you had eight principles already. What are the fundamentals?

Martin: Let’s say we’ve defined our principles. I’d say there are three fundamentals. There are the principles themselves. There’s the participatory aspect: we have sessions guided and facilitated by a STAIR Master, which is just a term for the person who knows these principles by heart and can guide and facilitate the conversation. And then we have a set of practical tools. In our case, a STAIR board — it’s just a Microsoft list where we register the conversations and the insights. We facilitate conversations using these boards and visualisations, but it’s really simple. It’s based around these sessions, these conversations, which can take the form of conversations or workshops, as you know from Agile.

Noz: What happens in one of these sessions?

Martin: Depending on the scope of the use case, if it’s larger, the STAIR Master prepares first. Sometimes there will be a brief describing the use case and its intention. You prepare in advance which of these principles are most in play for this particular use case. Then you facilitate the discussion as a workshop, asking the participants to reflect on those principles. The trick is being a facilitator and not putting people on the spot as if it’s an exam, but having them reflect: what’s the value proposition? What kind of value do we get? Do we have access to learning? Does this change relationships with our colleagues or within our own group? The STAIR Master’s job is to ensure the conversation stays within the sociotechnical framework.

Noz: So imagine I’m a leader with a team and I’m trying to implement this. Where do I begin?

Louise: I would say you begin by gathering people and telling them: today we’re going to use one hour to talk about and reflect on the impact of AI in our work, and also on the outcomes this team is obliged to deliver in the organisation. So you’re tied to strategy. Then you start with one use case — an AI model that is changing how you report or document or carry out tasks. The first question you ask is: what is the value that is added to our workflow, our present workflow? And then: what could be added by the new AI process? How is work changing? Is quality improving or is it actually decreasing? Talking about value in a broad way gives people the opportunity to weigh in on what value means to them. Because in a sociotechnical perspective, value includes humanistic objectives as well as economic objectives. And quite fast — because these sessions run maybe 30 minutes to an hour — we move to the five questions.

Noz: What are the five questions?

Louise: These are the five questions for the first step of a STAIR approach. The first question is: what is the value that you imagine or experience? It can be asked at any point in time — before an AI model is being introduced, or after a year and a half of using it. The second question is: what role does AI play in your work? You speak about the various ways of using it on a day-to-day basis. Is it an idea generator, or is it actually doing your reasoning? You discuss that in a psychologically safe environment, because this is your team leader asking how you use AI in your daily work. Question three is: what changes do you see in the long term? This is where people start opening up — both on the positive sides and on areas of concern. For example: “I don’t see as much of my colleagues as I used to” or “I worry about whether we trust each other’s outputs when we’ve used ChatGPT.” Question four is: how will you ensure wellbeing and efficiency in the team? That is the sociotechnical output. And question five is: can we describe the values that are important to us? What are the fundamental values you think about when you walk into work and deliver on a daily basis? Is it quality, wellbeing, being a good colleague? You distill those values from the discussion, and they go into describing what the set of principles will be that you then talk about in the open — in ongoing sessions, maybe once or twice a month. Because micro-changes to work are happening very fast with these new tools coming in. We need a legitimate way to talk about that.

Noz: So this provides a forum, and it also provides a conversation structure. Because I find it very difficult to have these conversations. Many people’s initial reaction is fear, and they can be the most vocal. The ones who actually get along with AI and feel comfortable with it sometimes don’t have much to say. But you’re giving them a structure that isn’t just about how to use the AI — it’s about that in a bigger context.

Martin: I think that’s right. I like the word intuition — we’re also building our intuition on AI. Both extremes, the very fearful and the very optimistic, present their own problems. What we need is the conversation in the middle, getting these different experiences and perspectives into play. What Louise was describing is the first step. The next step is the ongoing practice of reflection. The understanding here is that AI introduction and what comes with it isn’t a project with a fixed scope that reaches stability and then moves on. We can’t really do that anymore. This is a continuous reflection, a continuous adaptation to the technology. There’s no fixed destination, and the only real response is to keep reflecting.

Noz: I love that. The change is so big and so fast. You can’t have a pre-discussion. This is iterative. You have this reflection on an ongoing basis. I can’t come into a training course and make people understand how they’re really going to feel once this is put into practice. What we’re seeing is the difference between “this is just a new tool” and “this is actually going to reinvent how we work.” All of our processes are impacted. You need these conversations because you’re going to make incremental changes and those will have repercussions, and you need constant reflection. Can you walk us through an example? How did you apply the methodology and how did that go?

Louise: I’d like to give an example from a large hospital we worked with. They had an AI model being built and trained to take over some pre-surgical information for patients. Imagine you have a chatbot and a patient can ask it: how should I prepare for the operation next week? Very useful. We went into the department that was about to start using it and made interviews — asking nurses, medical secretaries, and surgeons how this would impact their everyday work. What would be the value? What role would it play in future processes? They started to think not just about the technology, but about how it changes decision processes and working modes. They could see that having the nurse not be the one reassuring the patient — that was chipping away at something important. A nurse not trusting an AI model to give a patient the right information about how to prepare before surgery. And in a hospital context, work is time-pressured, there is a lot of relational coordination between different professions, and the work is very technology-intensive. After going through the first step of STAIR, they formulated sociotechnical principles. Putting the three professional domains together — nurses, doctors, and medical secretaries — they found what was actually the core element that gives work meaning. When we introduced generative AI, we must maintain emotional communication in our workflows. That was one principle. The communications with the highest emotional importance are the ones we want to protect. And another was around training and experimentation. Because of the high level of real-time relational coordination at the hospital, protecting that was very important to them.

Noz: I think there are examples a lot of our listeners will recognise. Many will be in UX design, writing, technical writing, or product development. AI can replace your colleague interactions. Instead of having a reason to go down the hall or pick up the phone to ask a colleague something, you ask the AI. Yes, you’re more efficient as an individual. But you’re also just sitting at your desk all day talking to a machine. That can be damaging. And coders or writers, for example, have spent years developing a particular craft. They’ve got excellent grammar, excellent spelling, precise writing. They’re proud of that. Those are skills they worked hard to develop. And then you say: well, we just don’t need that anymore. That can be very shocking. It’s what I’m getting from your methodology — the conversation about what is the thing that is still valuable that I bring to the party. In one of our first Truth Collapse episodes, someone said we’ll probably have job security, but we won’t necessarily have job role security. Our job descriptions are being rewritten. That’s not a small thing, especially for a lot of people all at the same time, in ways no one could have predicted upfront.

Noz: I had a specific question for you, Martin. In the execution of this, what artefacts do you produce? Do you create a list of risks, notes about how roles are changing, acceptance criteria or metrics? What do you write down after one of these sessions, and how do you store that and come back to it?

Martin: We try to keep it very practical and at a minimum. We have a STAIR board — as I mentioned, just a Microsoft list, though it could be any programme. We keep track of what came up in the conversations. It becomes an idea bank. There’s an iterative, circling-back loop. Considerations we made, points we made, choices we made — we may want to change these further down the line. We can circle back using the STAIR board to look at what choices we made and whether the circumstances have changed. Those could be risks: technological risks, compliance issues, quality issues, ethical issues. Circumstances may change and we’ll want to revisit earlier decisions. That’s the most concrete artefact. The more abstract side is the ongoing reflection and the intuition we’re building — as STAIR Masters, but also as the professional groups participating in these sessions.

Noz: Are these forms available? Is the methodology published somewhere publicly?

Martin: There’s a webpage: stairmethod.org. The methodology is published there. There’s a guide, a questionnaire guide with the questions, and a set of pre-built principles you could use to get familiar with the thinking and the process. So there’s starter material you could just take. But if you want to go a step further, as Louise mentioned earlier, you’d work on your own principles, going through those five questions in the first step. That’s also available on the website. It’s all open source.

Louise: We’re actually working on designing a template where teams can document what AI models they’re working on. If you’ve asked several teams to conduct a STAIR session, how do you create transparency across those sessions? How do you share what one team is doing and what values they’ve identified, so other teams can learn from them? So team leaders, HR, or whoever has taken responsibility for rolling out AI in the organisation can have some element of documentation and transparency around who’s doing what.

Noz: From a practical perspective, who does what changes a lot. And from a social relations and meaning perspective — how teams feel like they are teams. If more of us are working from home and having less interaction anyway, there’s a tension between what technology is enabling us to do and how it gives us autonomy on one side, while we need to make sure that autonomy doesn’t tip into isolation.

Louise: That’s actually what I wrote my PhD about. But that’s a whole other conversation.

Noz: Maybe we’ll have you back for that one. How does it differ when applied to smaller versus larger initiatives? I can imagine a team of 10 or 20 getting in a room and working through this. But if I have a hundred people, or if I’m rolling this out globally across an enterprise — which is the kind of scale most of my clients work at — how do you scale this to larger groups?

Louise: We’re still working out the different levels. We are currently speaking with HR departments in large organisations who want to be part of this agenda. There are important decisions around workforce planning and work planning where HR can step in. The way we’re approaching it now is training senior HR professionals to take STAIR with them — to then train team leaders or other people leaders in the organisation who have responsibility for both wellbeing and rolling out new technology. I’m speaking with some large organisations about this. In a large organisation, it could be the people tech or people and culture team in HR who take on the responsibility of scaling STAIR.

Noz: My reaction to that: I’m not anti-HR. I think it’s valuable. But from what I’m seeing with clients, job roles are changing so fast and so profoundly that the business is having to reinvent the work on the fly, and then come back to HR to say: this is how we’ve restructured. There’s a positive feedback loop where HR trains business stakeholders on the methodology, but the business is moving fast and HR can feel behind the curve. I think things are often happening right where the work is being done, and the updates are going back to HR retrospectively.

Martin: I think the question is whether we should even frame STAIR as change management, or as a supplement to it. Change management as traditionally practised is planned, scoped, and aimed at reaching a point of stability. And within change management, there’s a perspective that change is ongoing — but the way we practise change management is very project-oriented. I wouldn’t call what STAIR describes “change management.” It’s more like ongoing transformation. HR has the position and the role in the organisation to be the agent of that continuous transformation. But you could apply the same critique to IT and digitalisation — they also work within project frameworks with an imagined endpoint of stability. That whole game has changed. We need agents who can drive continuous adaptation.

Noz: Yes. There’s a classic tension in organisations between shared functions like IT and HR, and the teams that are in their own area trying to define their own processes. We’re running low on time. This has been a very useful conversation. I think you’ve probably got some people who are going to visit your website to learn more. One final question for you both: what do you think the most important changes are that organisations should make in relation to AI and their approaches?

Martin: Redistribution of power.

Noz: I love that.

Louise: For me it’s acknowledging that technology drives change — but that change involves changing work roles, decision-making processes, and work systems. Acknowledging that, and then being very active about practising design, job crafting, and having a way of steering that. That’s really important.

Noz: I can’t agree with that more. I see so many people wanting to pour AI over the existing work, assuming nothing is really going to change. That the whole machine is just going to go faster. When actually the machine is fundamentally changing underneath. Martin, can you elaborate a little on what redistribution of power means to you?

Martin: I imagine the future organisation we’re talking about is a much more fluid organisation, where governance is much more fluid. Each of us has to take on more responsibilities. We’ve departmentalised governance, HR, and change management. I think all of that is going to become more distributed — more embedded within everyday work. The participatory aspect of this demands it. You can’t top-down steer this. You need the organisation to steer itself from within. That said, you still have an organisation that moves in a certain direction and makes choices. I’m very curious about what the future brings. I can’t imagine it retaining the same hierarchical structures organisations have today.

Noz: That makes me want to have a whole other conversation about the future of hierarchies, strategy, and leadership. Thank you both so much, Martin and Louise. We’re going to jump over to Truth Collapse now and talk about the more theoretical, philosophical, and sociotechnical dimensions of all this. I hope to see you all there. Remember to tell your friends — that’s our main form of awareness. Word of mouth. I personally hate saying “like and subscribe”, but I’m apparently a victim of the system.

Martin: Like and subscribe.

Noz: Thank you so much. Share, comment, tell your friends. Click all the things. Thank you both. I appreciate it, and we’ll see you again soon.

Martin: Bye.

Noz: Bye-bye.