To build an enterprise brain, connect your meetings, emails, tickets, and systems of record into one shared memory that remembers, reasons, and acts. Start with understanding, because understanding is what lets AI act on your behalf. Your company does not think today. It reacts. Every meeting gets recorded, every transcript gets filed, and every summary lands in an inbox nobody reads. The information exists. The insight does not. Note takers cannot get you there, because a summary sent to one person is knowledge buried in one more silo.
What is an enterprise brain?
An enterprise brain is a shared organizational memory that remembers, reasons, and acts across people, systems, and AI. It connects meetings, email, chat, documents, and systems of record into one governed source of understanding, then turns that understanding into outcomes. It replaces disconnected note takers and isolated knowledge bases with a single view your agents and your people can both query.
John Michelsen draws the line clearly. Understanding is the prerequisite. Understanding lets you reason. Reasoning lets you act. Skip the foundation, and your agentic strategy collapses, because an agent that does not know the company cannot act on its behalf.
Why do meeting note takers fail at enterprise scale?
Note takers fail because they capture knowledge and then bury it. Each one scopes to a single user, so a four-person vendor call ends up with three separate bots and three separate summaries of the same conversation. The output floods inboxes, never reaches the manager who needs visibility, and stays trapped in the silo it started in.
John is blunt about it.
“Those meeting summaries are spam,” he said.
You attended the meeting to hear it. You do not need an email puffing you up with what you already know. What you need is the one detail from three months ago, woven together with others, that surfaces the moment a problem appears.
The summaries also carry risk. Class action lawsuits target note takers that sent recordings to people who never agreed to be recorded. A side comment about a prospect, captured and forwarded, has burned real deals.
How is an enterprise brain different from a knowledge base?
A knowledge base stores answers. An enterprise brain acts on them. The difference is layered access and live connection. Information scopes to the right level, all-employees, department, or single project, the way security policies already work in LDAP and Azure. The brain pulls from systems of record in real time, so a closed JIRA ticket or a new HubSpot deal updates the understanding the moment it changes.
Chris Kraus frames the tribal knowledge problem by department. Support knows which features customers use. Sales knows what buyers want. Product knows the effort behind each release. Start with meetings to capture that intellectual property, then connect the systems of record so the brain sees how many tickets opened on a feature and whether anyone uses it.
A pile of recordings on a hard drive is not a brain. Real-time connection is.
What can an enterprise brain actually produce?
It can produce analysis no individual could assemble by hand. Michelsen described a recurring request from customers asking Krista to build an LLM access product. Instead of guessing, he asked his own reporting agent for the investment thesis.
The agent did not apply Reddit reasoning from a generic model. It drew on thousands of meetings, the full product roadmap, the change management system, and existing sales contracts, then layered in web search for competitive intel. It predicted how much of the product already existed in the architecture and how a launch would complement or compete with current offerings.
The result ran 80 pages, heavily cited, with internal and external references. “My first business plan was 30 pages,” Michelsen said. A startup could never have assembled this by hand.
How do you handle privacy when connecting everything?
You apply the access policies you already use. The cleanest approach extends the authorization and permission rules governing your corporate data to your meeting data. Be transparent, allow the opt-out, and keep audited policies for customer data. The failure pattern runs the opposite direction, where people pass links around for access and an employee violates a policy without knowing it.
This problem is not new. Companies have filled Teams chats, Slack channels, and inboxes for 30 years, and recorded meetings since Sony built the first mini recorder. The discipline already exists. The job is applying it.
Where should a company start?
Start where the work stitches together, usually the customer journey. Follow the path your customer takes through your organization, because that path crosses the departments you siloed and follows the dollars. Begin anywhere on it. A larger enterprise starts at the department level, then connects the fencing around each group.
The customer journey is not universal. A pharmaceutical manufacturer centers on research and R&D, so its brain belongs there. The rule is simpler than the example: whatever connects your key data sources and business processes is the right starting point.
The timeline is shorter than most expect. Chris puts conversation agent deployment at a couple of days. Within one to two weeks, knowledge flows into a central repository, and you can draw lines across meetings. People stop asking, “What did we decide last week?” In two weeks, your intellectual property becomes a knowledge base.
The real return is action, not another inbox
GenAI generates text. An enterprise brain generates outcomes. Organizations are cutting GenAI budgets right now because tokens fly across the internet with no measurable productivity to show for the bill. The fix is not more content. The fix is understanding that begets reasoning, reasoning that begets action, and action that runs at machine speed.
Krista is our brain.
One Krista customer put it plainly to John: “Krista is our brain.” Nothing happens in that business outside its view. That is the operating model worth building toward.
Stop filing summaries nobody reads. Build the understanding your business can act on.
Links and Resources
- Building a Cognitive Enterprise
- In Good Company with Nicolai Tangen podcast
- State of AI in the Enterprise, 2026. deloitte.com
Speakers
Transcription
Scott King
Welcome to this episode of the Union Podcast. We took a short break, and we are glad to be back. Hey Chris. Hey John, how are you both?
Chris Kraus
Doing well. Ready for an exciting conversation.
Scott King
Of course it is exciting, because we are talking about the enterprise brain. People describe it several ways, but they mean the same thing. They want an all-knowing place where they can query any type of information at any moment. I have heard it called a digital brain. We have used cognitive enterprise. Most people default to enterprise brain. John, I want to talk about what this means for enterprises with five hundred, two thousand, ten thousand employees. Not small companies. A lot of tribal knowledge gets lost. There is plenty of information, but they lack insight. What does it mean to you, and how do customers describe it?
John Michelsen
It matters to companies of all sizes. The tribal knowledge problem starts as soon as two people work together. One knows something the other does not, and that rift grows. The brain is far more than understanding, but understanding is the prerequisite. Understanding lets us reason. Reasoning lets us act. If your AI strategy is to go agentic without a basis in understanding, how will the agent work on behalf of a company it does not even know? We have to start with understanding, and it cannot be siloed, because siloed understanding causes our organizational problems today. We need a single view of the customer to know how an agent should react to a customer issue.
Then understanding leads to action, because we are in this for outcomes. If we only build knowledge bases and run tactical GenAI tool execution, we will not see ROI. As we record this, countless organizations are cutting their GenAI budgets. Tokens fly across the internet, but how much returns actual ROI? They cannot quantify the productivity. They can certainly see the bill. Understanding enables reasoning. Reasoning enables action. Those actions become the real return on investment. I do not mean actions by people. We already overwhelm people with information. The goal is for AI to hold the understanding, the reasoning, and the ability to act. That is when the business transforms.
Chris Kraus
When a development manager goes on holiday, the meetings get recorded and transcribed. But who has time to open each transcript and hunt for an action item? They need to ask, “In my absence, are there action items I need to cover that no one else handled?” When we release software, DevOps, QA, development, and product management all coordinate on one release. Visibility across even part of one department is hard. People think a list of action items means ROI. Is a list enough without understanding the bigger process?
John Michelsen
We are in the note taking business with our meeting agent, but frankly, those meeting summaries are spam. How often are you dying to see the summary from the meeting you just attended? You do not want your inbox flooded with them. I attended the meeting to hear it. The real problem is that three months from now, I need a kernel of information I did not recognize as important, woven together with others to solve a problem in that moment. Ideally an AI agent solves it on my behalf. Flooding inboxes with another note taker email is not the goal. We do not need another system that puffs us up with information. We need to bring information together.
Scott King
The risk can outweigh the reward. There are class action lawsuits against some note takers, because a summary went to a client who never agreed to be recorded. People talk after a call, sometimes about a prospect or customer, and that summary gets sent out and burns the deal.
John Michelsen
Think of the hot mic moments in politics. Now everyone gets to participate.
Scott King
You have to be careful. Architecturally, these note takers are individual. I was on a vendor call with four people and three different bots, because everyone signed up on their own. That will not work. It stays siloed.
Chris Kraus
Right. You get the summary you wanted, you send it out, and someone sends you theirs. You can end up with three summaries of one meeting. Just because you attended does not mean your team or your department has visibility. When I am on vacation, my manager needs to know what happened. No one can join every meeting, and you do not need information from every meeting, but you need access at the department and enterprise level. If you missed the all-hands on company benefits, you should be able to ask the enterprise brain, because that is scoped to all employees. HR policy changes are scoped to HR. Finance scopes its own. Consulting scopes customer one separately from customer two. There are layers, like security, LDAP, and Azure. The support agent needs project-level context on a support call. Customer success and development need visibility too. Silos do not help each other.
Scott King
On the Good Company Podcast, Reid Hoffman said this technology already exists and everyone is already recording meetings, but they get no insight because it is all disconnected. John, the obvious starting point is recording meetings. What else belongs in an enterprise brain? There is far more disconnected data in CRM and ERP systems.
John Michelsen
It certainly is not an individually scoped meeting bot. That is a note taker, not enterprise understanding. You build understanding across meetings, not your one-on-ones or private meetings, with the security measures any data stewardship requires. Then you integrate appropriately scoped emails, Teams and Slack chats, document repositories, and the systems themselves. We need to query those systems with the same understanding.
When you mention the Reid Hoffman point that no one gets insights, I beg to differ. I have shown you internally. Two or three times a month, a customer asks us to build a product around LLM access. I keep sending them to others in that space, and they come back saying that stuff is not any good and they want us to do it. So I go into our own reporting agent and ask what the investment thesis for building this product would be. Is that a prompt to a random GenAI model applying Reddit reasoning to our investment decision? The opposite. It draws on thousands of meetings, our entire roadmap and documentation, our change management system, and existing sales contracts.
It gave clear insight. It predicted how much of the product already exists in our architecture, how we message today, and how a new product would complement or compete with what we sell. It assessed whether competitors have a leg up, based on integrated web search and agentic analysis. So much of that thesis is based on who we are, what we do, and what our pipeline looks like. I ended up with an 80-page investment thesis, heavily cited, with internal and external references. My first business plan was 30 pages. A startup could never have imagined this.
Scott King
And that started because someone mentioned it would be cool. We run this enterprise brain internally, but not everyone has that capability yet. What is the path? If a CTO wants to run the same query, “Should I build this feature, deprecate this, or accelerate that, based on the roadmap and the sales forecast,” how do they get there?
Chris Kraus
The hard part is tribal knowledge. Each department holds its own. Support knows which features people use. Sales knows what buyers want. Product knows what is inside the product and the level of effort. All of that tribal knowledge comes out of meetings, decisions, and discussions. Start with meetings, because that is how you capture intellectual property. Then connect your systems of record. How many tickets opened on a feature? Do people use it? That lives in your ticketing system, your Salesforce, your JIRA. We plug ours into JIRA and our enterprise data, so we have more than meetings. You need real-time connections, because those systems change every day. It cannot be a pile of recordings on a hard drive. Every new HubSpot deal is a new opportunity. You connect across systems of record to get good insight.
Scott King
John, what are the privacy ramifications of connecting all this? How do you ease those fears?
John Michelsen
Whoever you partner with must be transparent, clear, and allow the opt-out. I am a privacy advocate and a cybersecurity professional by trade. Both vectors carry enormous consequences. People have recorded meetings since Sony built the first mini recorder in the seventies. Laws allow it in some circumstances and not others, so follow the rules. The best way, if you think organizationally, is to apply the authorization and access policies you already put on corporate data to the meeting data. Then keep conscious policies about customer data, which we already audit through a third party. Do it in the opposite order, where people pass links around for access, and an individual can violate a company policy without knowing it. We fill Teams chats, Slack channels, and inboxes every day, and we have done it for 30-plus years. This is not new. It is just easier now, so apply the data privacy and regulation discipline we already know.
Scott King
How long does this take for someone else to do? We built it, so we installed it and it worked. For a company already recording meetings with another tool, what is the timeframe to reach that understanding?
John Michelsen
We have customers who have done it. One of my proudest moments was hearing a customer CEO say, “Krista is our brain.” Nothing happens in that organization that does not run through Claude and Krista. Everything his people do runs through Claude into Krista, or Krista triggers people, Claude, or systems directly. Nothing in the business sits outside Krista’s view, and that is the basis of how they operate. That is a lighthouse account, and we think it should be the operating model for every business. A larger enterprise starts at the department level, then stitches together the cement-block fencing around each department. The customer journey is the right path for most businesses. Follow the path your customer takes, start anywhere on it, and use it to bring the organization together. Once you do, every part of the business runs faster and smarter.
Scott King
I am biased toward the customer path, because you follow the dollars. Sales brings them in, support reduces churn. Those organizations already write down procedures because of turnover, even if the procedures are not always current. Did your customer follow the customer journey, or start somewhere else?
John Michelsen
Ironically, they did not. Think of a pharmaceutical manufacturer. The customer journey is not the right path, because the business is so research and R&D centered. That is where the brain belongs. So it is not a comment for everybody. You need a path through your organization that stitches together what you tore apart. We deployed siloed SaaS, isolated the data sources, and gave each employee logins to only some systems. Teams at the largest companies steward who needs access to what data, and for how long. That is correct. But it flies in the face of needing enterprise understanding so the business runs at machine speed and AI makes intelligent decisions instead of half-informed guesses. Whatever stitches together your key data sources and business processes is the right path. For most it will be the customer journey.
Scott King
The FDA approval process runs for years, so for pharma that is following the dollars too. Chris, closing thoughts for someone who wants to try this? Where should they look, and how long should they expect it to take?
Chris Kraus
Implementing our conversation agent takes a couple of days. The trick is getting employees to move to one consistent, enterprise-grade conversation agent. After one week, then two, the knowledge starts flowing into a central repository, and you can draw lines across multiple meetings. You do not need months of data. People ask, “What did we decide last week? Two weeks ago we discussed this and I do not remember what we did.” Within two weeks you have more knowledge at your fingertips than before. In a meeting we can ask, “When we discussed this, what was our action plan? Was it implemented elsewhere? Why did we make that decision?” My question is always why. There was an architectural reason, but no one remembers it. In a couple of weeks of meetings, your intellectual property becomes a knowledge base.
Scott King
John, any closing thoughts?
John Michelsen
Here is a concrete example of a customer focused on the right things. They want to uplift their people’s productivity and manage without micromanaging. They collect the action items from every meeting, generate weekly summaries of what each person has and has not completed, set up next week’s tasks, and deliver pre-reads before important meetings so people remember what they should have accomplished. The executive thinking behind it: I want my people aligned to the goals they signed up for. I want them effective in meetings, because I am tired of meetings but we have to have them. And I need the understanding so that weeks or months later, I can see how something ended up the way it did. That was people-oriented productivity, built on understanding from many systems, not just transcripts. Understanding begets reasoning, which begets outcomes at machine speed. I am suspicious that a pile of note takers throwing emails into inboxes returns massive ROI. Weave that content into systems data and unstructured data with an AI fabric on top, and I see the opportunity to make money. I definitely see how enterprise understanding does.
Scott King
Thank you both for joining me. The enterprise brain is a recurring theme, and it will stay hot with us. Until next time.
