21 people, 4 teams. They adopted AI tools across the organization 3 months ago. Leadership is excited. Individual output is up. But something isn't adding up.
+21%
Code commits, PRs, messages per person
0%
Features shipped, cycle time, goals hit
5
Behavioral signals that need attention
Everyone is producing more. But the team isn't moving faster.
TeamWeaver connected to Meridian's Slack, GitHub, and Google Calendar. No surveys, no extra work. Here's what the communication patterns revealed.
Individual output keeps climbing while participation falls. Six weeks of data.
Six weeks of public-channel activity, read across the four teams. This is the surface Meridian's leads open on a Monday.
Signal snapshot · 4 teams · 6 signals · latest analysis
These are rates of observed behavior across public channels, read at team level. Nothing here is a grade and no one is compared to anyone else. Three things are worth your attention, and every reading behind them is in the portraits below.
Note 1 · Network Balance
Engineering reads 0.66, where teams typically sit at 0.35 or below. The other 3 teams run from 0.18 to 0.38.
When most paths run through a few hubs the team moves quickly, until one of those people is away. This is a resilience question more than a workload one.
Where to look
Pick one recurring thread and ask whether anyone else could pick it up tomorrow.
Note 2 · Thinking Diversity
Data Science reads 0.38, where teams typically sit at 0.65 or above. The other 3 teams run from 0.40 to 0.71.
Shared framing makes agreement quick and blind spots shared. Nothing here says the thinking is wrong, only that it is converging.
Where to look
Before the next decision, ask someone outside the team to read the plan cold.
Note 3 · Participation Balance
Engineering reads 0.58, where teams typically sit between 0.12 and 0.35. The other 3 teams run from 0.19 to 0.31.
Groups where a few people hold most of the speaking turns tend to think less well together than groups where turns are spread. Naming it early is easier than unwinding it later.
Where to look
In the next team meeting, ask the two people who spoke least what they would change.
Also in this analysis
Product and Design are the quiet ones, and that is worth saying out loud: nothing on any of them sits far from the range teams usually show. Worth a glance too: Engineering's trust language, Data Science's question rate and Data Science's cross-team reach also sit outside their typical ranges — they are in the portraits below.
Team portraits
The same readings the notes draw on, in full. On each rule the mark is where the reading sits and the shaded band is the range teams typically show.
Network balance, thinking diversity and three more sit outside their typical ranges.
Nothing here sits far from the range teams usually show.
Nothing here sits far from the range teams usually show.
Thinking diversity, trust language and three more sit outside their typical ranges.
Readings cover public-channel activity from the latest analysis, at team level only. There is no overall score and no ranking between teams. A snapshot shows you where to look, not who is ahead.
Sarah manages Engineering. This landed in her inbox before standup.
Here's what needs your attention on Engineering, Sarah.
Participation has declined 34% over 4 weeks. Two members haven't initiated a discussion in 12 days. Response latency up 2.3x.
Have 1:1 conversations focused on workload, not output. Consider redistributing the AI tooling migration that landed primarily on these two members.
78% of cross-functional messages flow through a single team member. Information bottleneck risk is elevated.
Introduce a rotating liaison role for cross-team coordination. Start with sprint planning.
AI-generated content rate is 31%. Cognitive diversity dropped to 0.38 and question frequency is down 45%.
Institute a 'challenge round' in design reviews where AI-generated proposals must be questioned before adoption.