Strategies for Encouraging Team Collaboration on AI Projects

Teams will increasingly include both humans and AI agents. We need to learn how best to configure them. A new Stanford University paper "ChatCollab: Exploring Collaboration Between Humans and AI Agents in Software Teams" reveals a range of useful insights. A few highlights: 💡 Human-AI Role Differentiation Fosters Collaboration. Assigning distinct roles to AI agents and humans in teams, such as CEO, Product Manager, and Developer, mirrors traditional team dynamics. This structure helps define responsibilities, ensures alignment with workflows, and allows humans to seamlessly integrate by adopting any role. This fosters a peer-like collaboration environment where humans can both guide and learn from AI agents. 🎯 Prompts Shape Team Interaction Styles. The configuration of AI agent prompts significantly influences collaboration dynamics. For example, emphasizing "asking for opinions" in prompts increased such interactions by 600%. This demonstrates that thoughtfully designed role-specific and behavioral prompts can fine-tune team dynamics, enabling targeted improvements in communication and decision-making efficiency. 🔄 Iterative Feedback Mechanisms Improve Team Performance. Human team members in roles such as clients or supervisors can provide real-time feedback to AI agents. This iterative process ensures agents refine their output, ask pertinent questions, and follow expected workflows. Such interaction not only improves project outcomes but also builds trust and adaptability in mixed teams. 🌟 Autonomy Balances Initiative and Dependence. ChatCollab’s AI agents exhibit autonomy by independently deciding when to act or wait based on their roles. For example, developers wait for PRDs before coding, avoiding redundant work. Ensuring that agents understand role-specific dependencies and workflows optimizes productivity while maintaining alignment with human expectations. 📊 Tailored Role Assignments Enhance Human Learning. Humans in teams can act as coaches, mentors, or peers to AI agents. This dynamic enables human participants to refine leadership and communication skills, while AI agents serve as practice partners or mentees. Configuring teams to simulate these dynamics provides dual benefits: skill development for humans and improved agent outputs through feedback. 🔍 Measurable Dynamics Enable Continuous Improvement. Collaboration analysis using frameworks like Bales’ Interaction Process reveals actionable patterns in human-AI interactions. For example, tracking increases in opinion-sharing and other key metrics allows iterative configuration and optimization of combined teams. 💬 Transparent Communication Channels Empower Humans. Using shared platforms like Slack for all human and AI interactions ensures transparency and inclusivity. Humans can easily observe agent reasoning and intervene when necessary, while agents remain responsive to human queries. Link to paper in comments.

The push for "AI fluency" can be intimidating for many people on our teams, especially those who have never written a line of code and aren't sure this is "for them...." I've had a lot of conversations with other leaders out there struggling with the growing gap between AI pros and generalists, so thought I'd share a few ways we're raising the floor and the ceiling at Doist, to bring our whole team from AI curious to AI fluency: → AI Lightning Talks — every two weeks, a Doister gives a ~20-minute demo of something they built or discovered, with live Q&A. → AI Office Hours — a recurring drop-in slot where anyone stuck, skeptical, or curious can bring a real problem and work through it with someone further along the AI curve. → Dedicated AI workshops at retreats — when we're together in person, we carve out structured time to learn and build together. Tip: offer parallel workshop options, one for newbies and another for pros. → An AI News thread — one place where the team posts the latest launches, articles, and tools worth paying attention to. → A "personal assistants" space — Doisters share the polished agents and automated workflows they've built for their roles. Usually frequented by the most AI-fluent on the team (raise the ceiling). → A generalist AI channel — a low-stakes zone for those earlier in their AI journey. Less advanced questions, prompts to push beyond your comfort zone, and a collection of wins for everyone (raise the floor). → And as we speak, we're in the middle of our biggest bet yet, "AI Week." We've set normal work aside for a full week so every team can operate with an AI-first mentality and every Doister can chase a side project that pushes AI further into how they actually work. Some experiments will flop, some projects will fall behind, but we've made peace with that and believe the learnings are worth more than the short-term output we're trading. I'll share what we learn in a follow-up post soon, so let me know if there's anything you want me to dig into when I do, and please share any other activities working for your team when it comes to bringing everyone along for the AI-ride 🌊

𝗛𝗼𝘄 𝗜 𝗴𝗲𝘁 𝗼𝘂𝘁 𝗼𝗳 𝗵𝘂𝘀𝗹𝘁𝗲 𝗮𝗻𝗱 𝘀𝘁𝗿𝘂𝗴g𝗹𝗲 𝗶𝗻 𝗽𝗿𝗼𝗷𝗲𝗰𝘁𝘀 𝘂𝘀𝗶𝗻𝗴 𝗔𝗜 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄𝘀 I’ve always worked on large corporate and consulting projects throughout my entire career. I can really say that I know the pain points in project workflows and collaboration. Project work is full of hidden friction: 🔄 Repetitive updates 🧩 Misaligned communication 📄 Documentation that never gets finished 🤯 Mental overload from managing everything Project work shouldn’t be this hard. I discovered that AI can be a game-changer. It’s a toolbox that quietly removes the friction, so teams can actually focus on creating value. 👉 Here are 3 AI workflows I can’t imagine project work without: 📊 Project Status Report Drafting 𝗣𝗿𝗼𝗯𝗹𝗲𝗺: Creating regular updates is repetitive and often delayed. 𝗔𝗜 𝗦𝗼𝗹𝘂𝘁𝗶𝗼𝗻: AI drafts weekly or monthly status reports from task data and notes. 𝗜𝗺𝗽𝗮𝗰𝘁 / 𝗕𝗲𝗻𝗲𝗳𝗶𝘁𝘀: Ensures consistent updates and professional formatting. 📍 Process Documentation Writer 𝗣𝗿𝗼𝗯𝗹𝗲𝗺: Documenting project workflows takes too long. 𝗔𝗜 𝗦𝗼𝗹𝘂𝘁𝗶𝗼𝗻: Converts bullet points into formal standard operating procedures. Rewrites complex content into plain simple language that everyone understands. 𝗜𝗺𝗽𝗮𝗰𝘁 / 𝗕𝗲𝗻𝗲𝗳𝗶𝘁𝘀: Supports scaling and standardisation. 👥 Meeting Summary and Clarification Generator 𝗣𝗿𝗼𝗯𝗹𝗲𝗺: Not everyone captures the same notes during meetings. Missing information or perspectives can lead to delays or conflicts. Hidden conflicts influence team collaboration in a bad way. 𝗔𝗜 𝗦𝗼𝗹𝘂𝘁𝗶𝗼𝗻: AI creates a neutral, complete summary including action items and decisions. Lists missing information, reveals hidden conflicts. 𝗜𝗺𝗽𝗮𝗰𝘁 / 𝗕𝗲𝗻𝗲𝗳𝗶𝘁𝘀: Ensures team alignment and saves time consolidating notes. Helps move forward faster and improves team collaboration by avoiding or solving conflicts. AI can really be a supporter for project teams, not replace them. And it is a true game-changer. I’m really happy to announce that Christoph Schmiedinger and I will start a content series about the practical usage of AI in project management and product management. We will keep you posted. Leave a comment about your experiences. Let’s learn together.

Most AI efforts don’t fail at the pilot stage. They fall apart after the pilot works. That’s usually where things go quiet. The problem isn’t the model. It’s that learning stays with one team, every new idea starts from zero, and momentum fades. Here’s the practical pattern I keep seeing in teams that actually get AI into production: 1) Put basic structure in place early (so people can move without fear) → Decide who owns AI decisions day to day. One clear point of accountability matters. → Set simple data rules: what’s allowed, what’s not, and what needs review. → Agree early on what “good” looks like: time saved, quality, cycle time, cost. → Keep governance light, but real enough that teams don’t hesitate. 2) Help more people get comfortable using AI in real work → Start with everyday tasks people already do: writing, summarizing, planning, analysis. → Teach by role, not by hype. Finance, ops, marketing, HR all need different examples. → Create a simple weekly habit: quick demos, what worked, what didn’t. → Build a small group of champions who help others without turning it into a big program. 3) Be intentional about which ideas get built next → Use a simple lens: impact, effort, risk, and reuse. → Prioritize the unglamorous wins that show value fast. They build trust. → Focus on use cases where data is accessible today, not “someday.” → Look for patterns you can reuse across teams: same workflows, connectors, and guardrails. 4) Treat deployment as ongoing, not a one-time launch → Build in steps: MVP, pilot, then broader rollout. → Measure as you go: quality, speed, cost, adoption, real business impact. → Get feedback early from users and subject matter experts. → Capture what works so the next build is faster and easier. A practical move you can make next quarter: Don’t ask teams for more AI ideas. Ask them which capabilities should be reusable across the business. That one shift changes how everything downstream gets built. If AI stays in pilots, it stays fragile. If it becomes a skill, it grows. If it becomes part of the system, it scales. 🔁 Repost if this helped you think differently about moving AI beyond pilots. 👤 Follow Gabriel Millien for practical insights on AI, operations, and turning ideas into execution.

We at Qwilr (like every other company right now!) held an AI adoption week a few months ago. Here’s how we made it actually useful/productive vs. performative… We had a genuinely excellent week - but ONLY because we thought carefully about the setup. 1. Make the mindset shift explicit, and lead it from the top This is a genuine reframe of how your team thinks about their work. If leadership doesn't signal that this is IMPORTANT, people won’t engage. Make expectations clear and enact them yourself. 2. Give people real space for self-discovery We gave everyone actual time (many people got most of a week, some got ~half of a week), away from normal work, to experiment. You cannot manufacture the "aha" moment for someone, they have to find it themselves. 3. Standardise the tooling (at least to start) Don't let 15 people use 15 different tools. Pick a stack, put everyone on it, reduce the friction. We (again, like everyone else on LinkedIn) are leaning hard into Claude & Notion. Give them good tools and pay for them properly so that people don’t hit token limits instantly. 4. Set up AI champions and office hours Have your most advanced people available to unblock others in real time. The difference between someone giving up on day two and someone having a breakthrough is often just having the right person nearby. 5. Require an AI project submission with a short (video) demo Non-negotiable for everyone (including me!). It creates accountability, surfaces incredible ideas, and makes the week feel real rather than theoretical. 6. Run internal and external talks Internal: who's already doing something interesting on your team? External: how are other companies actually approaching this? Both are genuinely valuable. Absolutely enormous shoutout to Steve Hind (Lorikeet , Mike Overell (ClassDojo) , Alex Hinds and David Parfett (Airtree) for giving up their time for some really incredible talks 7. End with a presentation/hackathon The demos at the end of our week were remarkable. Some of them were truly jaw-dropping. The team ended up building things we'd never have thought of, with new ways of looking at whole parts of the business. This is a fantastic way to wrap up your week. One honest caveat: a week like this will also surface a heap of REAL blocker stop adoption in your organisation, especially outside engineering. We were able to unblock some of them during the week - but even just discovering where those obstacles actually are is just as valuable as anything else. Would love to hear how other teams are running these.

An AI leader asked me how to improve the collaboration with business stakeholders: “I told them: This is what we do and this what we need from you. How can we get more buy-in?” I wasn’t too surprised that “the business” has been apprehensive if the tone has indeed been “us vs. them.” That’s why I suggested to reframe and rephrase it: “Our team understands the technology really well and we are looking to partner with you to uncover the most promising AI opportunities together, based on your domain expertise.” But that is just the first step. Next, we talked about governance. Creating a simple table of roles & responsibilities can already increase transparency and drive alignment (AI team | Business team). Add the technical and business roles you need to bring together to work on an AI idea, consider what each role brings to the project, and who meets with whom and how often. Building on that, we talked about the typical project phases from idea to operation to show the project flow. Add the deliverables and documents needed for each phase along with the outcomes and go/no-go criteria for the project. (Check out the chapter on building your idea funnel in the AI Leadership Handbook.) Lastly, we covered getting sign-off on this governance framework across your senior business stakeholders. This will set you up for an aligned approach with top-down support and help you shine in your AI leadership role. I’ll check in again in a few weeks and can’t wait to hear how things are going. What’s slowing your AI program down? (Drop me a DM for an unbiased perspective.) #ArtificialIntelligence #GenerativeAI #IntelligenceBriefing

Your role changes when your AI teammates start working together. Most teams brief each Custom GPT teammate separately and lose context in handoffs. Trailblazing teams connect AI teammates so expertise flows seamlessly from positioning to content to campaigns. One conversation, full context. Harvard research with P&G professionals shows that when people work with AI, "you stop caring as much about the normal boundaries of your job." Connected AI teammates speed up this transformation. Your positioning expert's GPT works directly with your content expert's GPT. Knowledge flows where customers need it, not where org charts say it should. This week's newsletter explores: ► Why most teams get stuck using AI individually instead of as connected systems ► How to build your first AI chain that combines multiple areas of expertise ► The three phases of AI adoption and why Phase 3 transforms your role from doing tasks to strategic work and orchestrating AI systems ► Real examples of teams rethinking work around customer outcomes, not org charts ► How a GPT Navigator helps you pick the right teammates for any project Teams connecting their AI are working differently. They eliminate handoffs instead of managing them while discovering their jobs are evolving in the process. Work is shifting around what customers need, not what our org charts say. Chained GPTs show you what that future looks like. Read the full issue below. There's also a 16-minute AI podcast version in the comments for those who prefer to listen while multitasking. See link in the comments. Huge thanks to Angie Hill (SVP of Growth and Integrated Marketing at Procore Technologies) and Maggie Miller (Senior Director of Corporate Marketing at HackerOne) for sharing how chaining AI teammates is changing their approach to collaboration and strategic work. The infrastructure is here. Will you keep working with AI teammates individually or start chaining them together? Share this with your team and others to inspire them with this vision and approach to transformation.