In November 2007, the collaboration stack at a typical mid-sized company consisted of four things: an email server, a shared network drive, a conference bridge, and whatever instant messenger employees had installed without asking. That was it. Microsoft had just released Office Communications Server 2007, bringing presence, instant messaging and voice into a single enterprise product. IBM had Lotus Sametime. Cisco had bought WebEx earlier that year in a deal reported at roughly $3.2 billion. The pieces of modern collaboration existed. What did not exist was the thing Slack would later make obvious: a persistent, searchable, shared place where work conversation accumulates. So everything went into email. Understanding why that failed — and what actually replaced it — explains most of what is wrong with collaboration stacks today.
Email as the Universal System
Email in 2007 was not a messaging tool. It was the workflow engine, the file repository, the task tracker, the approval system, the knowledge base and the audit trail, because nothing else in the building could perform those jobs. Each of those uses failed in a characteristic way. As a file store, it created version chaos. Document_v3_final_JS_revised.doc existed in eleven inboxes in four states. The authoritative version was whichever one the loudest person had open. As a workflow engine, it had no state. Whether an approval had been granted was determined by reading a thread and inferring intent. Nothing could be queried, reported or escalated automatically. As a knowledge base, it was structurally private. This was the decisive flaw. Every answer anyone gave to any question was delivered into individual mailboxes and disappeared. A new joiner could not find it. A colleague in another office could not find it. The person who wrote it could not find it eighteen months later. Organizational knowledge was being generated continuously and filed into hundreds of sealed containers. As a communication channel, it made copying people the only way to include them. Reply-all was not a social failing; it was the sole available mechanism for making a conversation visible to a group. The alternative was that the group never knew.
The Shadow IM Problem
Meanwhile, actual conversation had already moved. Employees were using consumer messengers to coordinate work because it was faster than email and their colleagues were there. From a governance perspective, this was the same pattern that produces shadow IT in every era: unsanctioned tools, no retention, no logging, no discovery capability, and business decisions being made in a channel the organization could not see. Enterprise IM products like Office Communications Server and Sametime were built substantially in response to this — not to enable messaging, but to bring the messaging that was already happening back inside the perimeter. They largely succeeded at the control objective and largely failed at the collaboration one, for a reason that took years to articulate: enterprise IM copied the consumer model of one-to-one and ad hoc group chats. The conversation was still ephemeral, still private to its participants, still unsearchable by anyone outside it. It solved the compliance problem while reproducing the knowledge-loss problem exactly.
Each event links to its supporting source. This is a selective chronology, not a performance comparison.
What Was Actually Missing
The insight that arrived later — Yammer in 2008, Slack from 2013 — was not chat. Chat had existed for decades. It was three specific design decisions:
- Conversation organised by topic, not by recipient list. A channel is a place, not a distribution list. People join it because the subject is relevant to them, not because someone remembered to include them.
- Open by default. Anyone in the organization can read most channels. Knowledge stops being private property and becomes searchable infrastructure.
- Persistent and indexed. The conversation survives the participants. A question answered in March is findable in November by someone who was not there. Together these converted conversation from a transmission event into an organizational asset. That is the whole innovation. Everything else — emoji reactions, integrations, threads — is refinement.
And Now the Pendulum Has Swung Back
The modern stack has the opposite problem. A typical company now runs a chat platform, email, a document suite, a project tool, a ticketing system, a CRM, a wiki and a video platform. Every one has its own conversation surface. The knowledge that used to be trapped in individual mailboxes is now scattered across eight systems, which is a different failure with an identical outcome: nobody can find the decision. This matters more than it used to, because AI assistants are now the primary way people search internal information. An assistant can only retrieve what was written down somewhere it can reach. If your decisions live in chat threads, meeting recordings and private messages, your retrieval quality will be poor regardless of which model you buy. Content placement has quietly become an AI infrastructure decision.
Designing a Stack That Holds Knowledge
- Write down what belongs where. Chat for coordination, documents for decisions and durable knowledge, tickets for tracked work, email for external parties. One page, agreed by leadership, referenced in onboarding.
- Default to open channels. Private messages should be the exception, and people should be able to say why a conversation is private.
- Require a written decision record. If something was agreed in a meeting or a thread, it does not exist until it is in a document with a date and an owner. This is the single highest-return collaboration practice available.
- Count your surfaces. List every place a work conversation can occur. If the number exceeds five, consolidation will save more time than any productivity tool you could add.
- Set retention deliberately. Indefinite chat retention is a discovery liability; short retention destroys institutional memory. Choose, document it, and make sure decisions are recorded somewhere permanent before the window closes.
- Check what your AI assistant can see. If it cannot reach the systems holding your decisions, you are paying for a search tool pointed at the wrong shelf. The pre-Slack era lost knowledge because conversation was private. The current era loses it because conversation is fragmented. The underlying question has not changed in twenty years: when this exchange is over, where will the answer live, and who will be able to find it?
Common Questions
Why did email fail as a collaboration tool?
Because it is structurally private. Every answer is delivered to specific inboxes and becomes invisible to everyone else, so the organization regenerates the same knowledge repeatedly.
What did Slack actually change?
It made work conversation topic-based, open by default and permanently searchable. Chat itself was decades old; persistence and openness were the innovation.
How many collaboration tools should a company run?
Few enough that any employee can say, without hesitation, where a given type of content belongs. In practice that means a chat platform, a document workspace, a work tracker and email — with clear boundaries between them.
How do we stop decisions disappearing into chat?
Require a written decision record with a date and an owner in a durable location, and treat the chat thread as the discussion rather than the outcome. Enforce it in the same review rhythm you use for everything else.
Assess Your Collaboration Stack — Outpace maps every surface where your work conversations happen, identifies where decisions and knowledge are being lost, and defines the content placement rules that make both people and AI assistants able to find them.