The Context Warehouse

One of the most frequent complaints I hear from people is that they have too many meetings, endless video calls, and constant interruptions from Slack. But when I ask people whether they have a good understanding about what’s going on inside their org, more often than not, they don’t.

Why can we find any random fact in one second on Google, but for most people, finding out basic company info like ‘What projects are being worked on?’ and ‘Are they on track?’ requires hunting for (out of date) docs, sending Slack messages, or attending lengthy meetings.

Often, the info you need does exist, but it’s hard to find. Sometimes the consequences of this are large. E.g. A project I know of at Amazon where employees spent months on two identical projects. But often, the cost is subtle and insidious: More time in meetings, more time on comms, a slower pace, and weaker alignment.

This is a public good failure. Context, widely shared, is (with a few exceptions) a positive, leading to greater efficiency, and better decision making, but individual teams are not strongly incentivized to share info beyond their immediate stakeholders. In the few companies that do this well (Shopify, Gitlab etc), leadership has mandated a certain way of working to overcome this public good problem, however they must also be careful not to constrain too strongly how teams work.

In general, a team’s ability to operate autonomously relies on alignment, which is strengthened by context. So low context companies, which rely on top-down comms and meetings to share info, will have lower decoupling, lower autonomy, and lower execution speed.

In contrast, I believe companies of the future will be high context, moving from a top-down, push model, to a bottom-up, pull model of information sharing. Context on demand.

At the moment, building a high context org relies on an opinionated leadership (to overcome the public good problem) alongside an async culture, and consistent tooling to support durable, discoverable, context artefacts (project briefs, updates, discussions, decisions etc). However, AI will provide another solution -  a way to move towards a high context setup, without requiring significant behaviour change.

Once costs have fallen, companies will be able to layer AI on top of their existing tools (Zoom, Slack, Notion etc), to store, auto-summarise, and organise all information generated by the company. E.g. The transcripts from every video call, every slack message, tasks, and contents of every doc will all be put into one data warehouse, or rather, a Context Warehouse.

Employees will be able to interrogate a context warehouse with questions like:

• Show me a summary of all active projects including whether they’re on track
• Which projects have unclear goals?
• Give me a summary of the thinking behind project X
• What are the common characteristics of previous projects which failed, and which active projects most closely match?
• Which projects contribute the least to our business strategy?
• Which recurring meetings provide the lowest value?
• Which teams exhibit the least psychological safety?

Companies building integrated products (like Microsoft) will be best placed to implement this kind of solution, however, I also expect other companies to provide this as a service, by ingesting data from popular tools.

Employees will be able to access information quickly and asynchronously, with a corresponding reduction in comms, and increase in efficiency.

In effect, AI will help all companies realise some of the benefits of async working, even if they continue to operate synchronously. This should be a win for everyone, even if true async companies will still benefit from wider talent pools and greater efficiency.