As AI moves from individual use into meetings, workshops, strategy sessions, and collaborative writing, it increasingly enters the room with us. Some find this useful, while others experience unease, often voiced in terms of the tools themselves. Looked at more closely, that unease tends to centre on authorship, responsibility, trust, and how collective judgement remains legitimate when AI-generated material becomes part of shared work. This post focuses on those practice-level tensions and what they mean for working together well.
This is the third in a short series of reflections on using AI in professional practice. The first post explored AI as a thought partner, a tool that can support reflection without replacing judgement. The second offered a simple prompt structure for using AI in ways that protect voice and ownership. This post moves the focus outward, from individual use to what happens when AI enters shared work.
Across teams, organisations, and facilitation settings, AI is beginning to be used in meetings, collaborative writing, and planning processes. Experiences are mixed. Some find it useful or productive in shared work, while others express unease when it enters collective settings. Where that unease is present, it is rarely about accuracy or efficiency. It tends to surface instead around authorship, responsibility, trust, and how judgement is exercised together, particularly when AI-generated material enters shared work.
These questions are not easily resolved through formal policy or agreed conventions alone. They signal that familiar practices of judgement and shared sense-making are being stretched. Paying attention to where discomfort shows up can help teams and facilitators work with AI in ways that remain supportive, rather than quietly authoritative.
The tensions people are noticing do not all show up in the same way. They surface in different moments of shared work, sometimes quietly, sometimes through friction or discomfort. What follows looks more closely at four places where AI’s presence in the room tends to make itself felt: how authorship is experienced, how responsibility is carried, how trust is sustained in shared spaces, and how convenience can slowly erode engagement. These are not separate problems to be solved, but overlapping practice challenges that shape whether collective judgement remains credible over time.
When help becomes authorship
One of the first places unease shows up is around authorship, particularly as AI moves from individual use into shared work. When people work with AI on their own, it is often treated as a drafting or structuring aid, and the boundary feels relatively clear. In collaborative settings, that boundary becomes harder to hold. Strategy documents and shared outputs are rarely owned by a single person, and AI can blur an already negotiated process.
Practitioners often describe uncertainty about whether work shaped with AI still feels like something they can stand behind. This is especially visible in writing-heavy roles, where contribution is closely tied to voice, craft, and judgement. When AI is used to restructure or polish shared documents, some people experience a subtle loss of recognition for their contribution, even when the content itself remains sound.
Working with this tension means taking authorship seriously as a practice rather than a label. In shared work, authorship needs to be shaped over time and actively negotiated, but in practice it often is not. Roles, hierarchies, and established conventions frequently stand in for open discussion of contribution. Introducing AI into this mix can intensify existing tensions, and in some cases create new ones, particularly for those who are already uneasy about AI’s presence in collaborative work.
Rather than trying to draw hard lines around what “counts” as human or AI-assisted work, teams may be better served by slowing down moments where authorship matters most. This includes points where work is shared externally, where commitments are being made, or where people are expected to stand behind particular claims. Naming how a piece of work came together, including the role of tools, can help restore a sense of agency and mutual recognition without turning transparency into surveillance.
What matters is not purity of authorship, but whether people still feel able to say, with confidence, “this reflects my judgement” or “this reflects our shared thinking.” When that claim remains intact, AI can sit in the background as support rather than becoming an unspoken co-author.
Responsibility without grip
A second tension shows up around responsibility. Across leadership, management, and advisory roles, there is broad recognition that AI can generate options, summaries, and recommendations quickly, but cannot hold responsibility for them. In principle, accountability remains human. In practice, that clarity often feels thin.
What people describe instead is a gap between formal responsibility and felt ownership of decisions. AI makes it easy to surface plausible answers faster than individuals or teams can fully work through the judgement involved, particularly under time pressure. In strategy and planning contexts, AI-generated frameworks or draft plans often enter the room already appearing complete, leaving limited space for shared sense-making. While often intended as preparation, this can feel to others like the difficult work of thinking together has already been done elsewhere.
Over time, this creates unease. People are effectively asked to endorse outcomes they did not meaningfully shape, or to stand behind recommendations that feel imported rather than worked through. When things go wrong, accountability snaps back sharply onto individuals, reinforcing the sense that responsibility has been shifted without being shared. Working with this tension requires distinguishing responsibility from participation. Responsibility in collective work is not only about being answerable after the fact, but about having had a hand in shaping the judgement that led to an outcome.
This work does not take place in a vacuum. Time pressure, organisational incentives, accountability regimes, and performance expectations often limit opportunities to slow down, reopen judgement, or create space for shared sense-making, even when people recognise the need to do so. Within those constraints, teams may need to pay particular attention to moments where judgement is re-entered. Slowing down adoption, reopening assumptions, or asking what has not been considered can help restore agency before decisions harden. When people can say, “I understand how we arrived here,” responsibility feels shared and credible rather than procedural.
Trust in shared spaces
Trust is often tested when AI outputs enter shared spaces such as meetings, workshops, and facilitated sessions. People express mixed feelings about whether and how to disclose AI use in these settings. Bringing AI-generated summaries or slides into a meeting can feel risky if errors or misinterpretations are exposed. At the same time, not naming AI’s role can feel deceptive, particularly in contexts that depend on openness and credibility.
This tension becomes sharp around records of collective work. AI-summarised meetings, action lists, or reflections often appear tidy and authoritative. Yet participants frequently report feeling misrepresented. Nuance, disagreement, and relational dynamics are easily flattened. When that happens, it becomes harder to contest the record or revisit decisions, especially if the summary is treated as neutral or final.
For facilitators and process-holders, this raises questions of legitimacy. Shared spaces rely on participants trusting that their contributions are heard, represented with care, and open to challenge. When AI quietly shapes what is captured and remembered, it can shift narrative power in ways that are difficult to notice.
Working with this tension means treating shared records as provisional rather than authoritative. Long before AI, notes and summaries shaped what groups remembered and acted on, sometimes unevenly and with contested effects. AI does not change that dynamic, but it amplifies it by producing outputs that sound confident and complete.
Rather than asking whether AI-generated records are accurate, a more useful practice question is whether they remain contestable. Making it explicit that summaries are starting points, inviting correction, and keeping ownership of interpretation clearly human helps protect the integrity of shared spaces. Trust is sustained not by perfect capture, but by the ongoing ability to question and revise what has been recorded.
Erosion through convenience
A final tension runs through many of the concerns above: erosion through convenience. Managers describe frustration when teams default to AI as a first step rather than a support for thinking. Drafts are produced quickly, often lacking context or care, and treated as “good enough”, narrowing opportunities for reflection.
Individuals express a related unease. The concern is not sudden deskilling, but a gradual thinning of engagement. When AI removes friction too effectively, it can also remove the effort through which judgement is practised and confidence is built. Over time, people worry about losing touch with their own voice.
These dynamics are unevenly felt. Differences in comfort with AI create new forms of team friction. Some people feel pressured to adopt tools that do not sit well with their professional identity. Others feel constrained by colleagues who resist AI use entirely. Without shared norms, these differences can harden into persistent tensions.
At the far edge of this concern sit anxieties about persuasion. When AI-generated insights are used to steer discussions or justify pre-set decisions, particularly without transparency, people feel their capacity for informed judgement has been undermined. In these moments, the issue is not efficiency but the legitimacy of collective decision-making.
Working with this tension requires distinguishing convenience from care. Convenience becomes problematic when it replaces engagement rather than supporting it. In collective work, some forms of effort are not inefficiencies to be removed, but sources of meaning, learning, and trust.
Rather than asking how much AI is too much, teams may find it more helpful to ask where effort still matters. Writing, discussion, and reflection are practices through which people test ideas and surface disagreement. Protecting space for that work helps ensure that AI remains a support for judgement rather than a substitute for it.
Concluding thoughts
Taken together, these tensions point to a common underlying issue. AI does not introduce entirely new problems into collective work. Instead, it amplifies long-standing questions about authorship, responsibility, trust, and how judgement is exercised together. What feels unsettling is often not the presence of the technology itself, but the speed and subtlety with which it reshapes practices that were previously held through habit, craft, and relationship.
Seen this way, the current moment invites greater deliberateness rather than firmer rules. In shared work, judgement has always depended on people having time and space to think together, to contest interpretations, and to stand behind what is produced. Used well, AI can sometimes widen that space, helping groups surface options, perspectives, or connections they might not otherwise reach. It also makes it easier to move quickly past those moments. It also makes visible where they matter most. Paying attention to discomfort can help teams notice when care, deliberation, or shared sense-making are being compressed too far.
Used with this awareness, AI does not need to become an authority in the room. It can remain a background support, useful but contained, while authorship, responsibility, and judgement stay clearly human and collectively held. Holding that balance is not something to optimise and move on from. It is an ongoing practice, shaped by context, relationships, and the kind of work people are trying to do together. This focus on practice does not exhaust the wider questions raised by AI, including its environmental impacts and broader societal risks, which matter but call for a different kind of reflection.
More information on the use of AI can be found through the LfS generative AI landing page. This includes links to the related resource pages Using AI in research and practice and AI in context: the wider picture. Also see the accompanying posts – AI as a thought partner: reflections on collaborative practice and systems work and AI prompts for shared thinking: a light framework for purposeful prompting. Collectively these pages introduce practical tools, ethical guidance, and curated articles to support thoughtful use of AI in business, applied research, and collaborative practice.
[* Image: wollyvonwolleroy | Pixabay]
