Getting Started
The Simple Logic
When you pick a plan, a swarm of 5 to 15 agents, and dozens of agent instances are immediately available to you and your team. You configure each agent's context and prompts to tailor their performance according to your needs. Rapid onboarding enables you to get up and running within hours. You don't need to be perfect since the agents will improve through interaction.
For less than the cost of an equivalent full-time employee or part-time consultant you then have over a dozen advisors and assistants working for you. Each month they unearth thousands of unique and applicable insights relevant to your organization and its ecosystem. They digest, process, and store this information, and make it available through always-on advisors that you can interact with any time. This empowers you to discover issues and opportunities, disseminate and align pertinent strategic knowledge and understanding, and level up your corporate cognitive and decision-making capacities.
What Do These Agents Do?
The agents keep the finger on the pulse of your organization. In essence, we have two kinds of agents, each with sub-categories: Assistants collect data or distribute knowledge in automated fashion, e.g. via targeted research output and advisors. Advisors are to be interactive agents that participate in calls or talk to you.
ADVISE: Agent Experts listen or participate in meetings, brainstorming, or analysis sessions. They become your own Panel of Experts. Doing so, these experts store transcripts, and also summarize and vectorize them for quick retrieval and improved context. They remember and they learn.
HEALTH CHECK: Agent Assessors perform regular check-ins, un-obtrusively. Through a quick talk, they informally hear about your daily or weekly progress, and about ideas and opportunities, issues and concerns. They digest and incorporate this in their knowledge base as well.
RESEARCH: Agent Assistants research data specifically relevant to your organization, about technology, markets, competitors, technology, and the macro context. They summarize this knowledge and distribute per your instructions. It not only informs you but also results in the agents improving since they store findings in postgres and vector databases.
The longer this continues, the better the agent's output can become. Happening organically and without disrupting existing workflows, it allows you to keep the finger on the pulse of your organization. More even, it turns these distributed uses of agents into both sensors that measure, and neurons that keep learning and improving. Since they are connected, information gathered can be instantaneously available to any other agent.
They become a corporate neural net, and evolve into a Corporate Cognitive Twin.
Where Does The Data Come From?
The agents acquire knowledge internally or externally. Some listen and talk. Others conduct research on external sources. They digest news, newsletters, podcasts, and conference presentations. They create and access transcripts from conversations and check-ins. They can scan email accounts. They also can use the Cognion platform to integrate with unique internal or external data sources or device sensors using APIs.
And so the agents learn, through formally modified fine-tuning and customization, but above all when they interact with humans. The more your team has agents listen or participate in conversations, the more check-in input you have given, the more relevant research "feeding agents" have done - the more relevant their output becomes. But you don't need to change your job: the agents work in the background, listen, or you talk to them like with any other team members.
Even if you were not giving them more formally customize (fine-tune) them, they would learn.
A Corporate Cognitive Twin Emerges
Since all knowledge is immediately made available, accessible 24x7, and across the organization via Advisors, they act as a cognitive layer establishing a constantly improving Corporate Cognitive Twin (CCT). Not limited by artificial borders of departments or even organizations, nor silos of information and echo chambers, Advisors add perspectives and further the spreading of high-value knowledge and thinking. They can discover non-linear relationships and deliver unmatched strategic advice.
Taking it From There
Through the just described process, the institutional knowledge of the organization gets gradually transferred to a network of "neurons" with their vector database and supporting data storage. What ERP systems gave you about numbers, your new Neural Net gives you about all the softer, more messy, but arguably more important matters in your organization: one source of "truth" and relevant knowledge. When you now integrate or automate processes, they are built on top of this corporate cognitive platform.
In the advanced case, this enhanced learning is then integrated into a custom-trained Small Language Model (SLM). This is your own Corporate Cognitive Twin model, trained on yourself, complementing the LLMs that are trained on the general knowledge of humanity. And slowly but surely, you have turned your organization into an AI-First entity.
(1) Neural Net Top Level
The above graph depicts how agents draw both on (a) their own specialized (agent-driven) sources of data and knowledge, but also (b) the knowledge accumulated via parallel specialized processes. The interaction with expert agents focuses on these agents' specialities and is driven by corresponding priorities and context. E.g., how are we defining the market, who are our competitors, which technologies and supply chains are crucial for us, and what are our corporate objectives or key results?
But since we are talking about a self-learning Neural Net, a Market Expert (agent) is not limited to that field. Rather, they can draw on additional context about what has been happening inside the organization. Input given in one part, a conversation taking place somewhere else, a technology development, issues that surfaced, all this is readily available to anybody that matters. Information flows freely, rapidly, and organically . Borders and silos have been removed completely.
(2) Neural Net Layered View
Below the level of expert interaction, we can see how agents access other special fields, beyond the listening or talking in conversations and talks. Each interactive agent can draw on data and knowledge extracted by agents focusing on other fields or specialized tasks. Direct Agent-to-Agent (A2A) interaction is not frequently needed inside an organization since the data can be shared directly.
However, A2A is feasible and technically supported by the Cognion platform. It then extends the reach into customer, supplier, and partner networks. The above depicts a way to integrate the larger ecosystem, in the form of customers and suppliers. Apart from authorization, there is no inherent limitation restricting agent interactions to one organization.
The Customer always has a seat at the table
One of the agents every organization should use is - the customer. It enables the "configuration" and definition of a customer persona as an advisor, always available to give the perspective of those that an entity serves.
Like a Real… Neural Network
Like a real neural network, all this data can be shared. When you ask a market expert how recent market developments affect a specific topic discussed, they are more than experts in market research. They also can draw on key points from conversations that took place in other parts of an organization, on the latest technology or macro-economic trends, or on ideas or issues reported by teams far away. The Market Expert will look at market data first, consider specific goals and priorities, but be aware of everythying else.
