This summer, we proudly sponsored the 20th Annual CDOIQ Symposium in Boston, USA. We led two sessions on an increasingly important idea: Swarm Intelligence in AI.
In enterprise AI, Swarm Intelligence means multiple AI agents that can (1) reason, (2) coordinate and (3) act together on using organisational data. Rather than working as separate assistants on isolated tasks, these agents can connect across teams and systems to support (and eventually carry out) processes throughout the organisation.
This swarm creates enormous potential for operating at greater speed and scale. But it also raises the stakes. The more autonomy given to AI, the more important the trusted data, business context and strong governance become..
More agents do not automatically create better intelligence. If they don’t understand how your organisation really works – its definitions, rules, relationships and processes – they risk becoming disconnected, acting on averages and producing decisions that are difficult to verify.
Across our two CDOIQ sessions, we explored both sides of this idea:
- what Swarm Intelligence could mean for organisations and the people within them,
- and the data and knowledge foundations needed to make AI execution safe, measurable and useful.
“Swarm intelligence: multiple AI agents that reason, coordinate, and act together on enterprise data.”
What is Swarm Intelligence?
Much of the first wave of generative AI has been about interaction. We ask a question > AI generates an answer. Agentic AI changes that relationship. An agent can be given a goal, make decisions about how to achieve it and execute tasks on our behalf.
Connect multiple agents together and the proposition changes again. Different agents can take on different activities, communicate with one another and move a wider process forward – bringing humans in where judgement is needed.
That is a very different proposition from giving every employee a chatbot. It raises a bigger question: what happens to the organisation when AI starts doing rather than simply answering?
Why does AI make dashboard prototyping faster?
This was the focus of our first CDOIQ session, Swarm Intelligence: From Orchestration to Organisational Impact and the Future of Work. Together with Dr. Hema Seshadri, AI & Data Engineer at Akamai Technologies and Kishore Aradhya, Data & AI Initaitives Lead, formerly at Frontdoor Inc, we explored how connected AI agents could affect organisational structures, skills, leadership and the relationship between augmentation and replacement.
As execution becomes faster and cheaper, human value shifts. Judgement, validation, critical thinking and the ability to understand context become more important, not less. The opportunity is not to leave people sitting above agents approving whatever they produce, but to use AI to create more room for work that requires challenge, nuance and accountability.
“AI, for me personally, has given me more opportunity to do more critical thinking."
Change management cannot be an afterthought
Technology is only one part of today’s AI transition. During the session, we asked the audience which human-related challenge mattered most when adopting AI. Of 42 respondents, 21 selected change management, compared with 14 for learning and education and 10 for fear of job loss.
Organisations can build sophisticated AI capabilities, but people still need to understand why they are being introduced, how to interpret the results, how roles will change and where human judgement remains essential. Swarm Intelligence is not simply an AI architecture question. It is an organisational one.
Watch the full panel discussion:
More agents do not automatically mean better decisions
Before giving agents more responsibility, we need to think seriously about what they are reasoning with. An AI model can be incredibly capable without knowing anything meaningful about the organisation.
It doesn’t inherently know what ‘customer’ means across your business, which policies apply to which processes, or why one rule takes precedence over another. Without that enterprise memory and context, AI is detecting statistical patterns in information it does not really understand. Access to information doesn’t result in intelligent decision-making.
With one AI assistant, that is a problem. With a swarm of agents, it’s a problem at scale.
Your competitive advantage is invisible to AI
AI models are increasingly available to everyone. As the model layer becomes more interchangeable, simply having access to powerful AI becomes less of a differentiator.
What differentiates your organisation is everything the model does not automatically know: your rules, definitions, processes, relationships and the knowledge your people have accumulated about how things actually get done. Much of that knowledge still lives in documents, systems and people’s heads rather than in formats machines can reliably reason over.
That is the gap many of our clients are concerned about right now.
“Your competitive advantage will not come from the model. It comes from the meaning the model can access.”
From managing data to architecting knowledge
For decades, data architecture has helped organisations answer an important question: what does the data say? AI introduces another: what does the data mean?
As an example, let’s use the word ‘customer’. A glossary can define the term, but your specific business meaning also sits in relationships and rules: how a customer relates to a product, what makes them eligible, which consent they have given and which policies govern how their data can be used.
That is why we believe organisations need to move towards knowledge architecture: making business meaning and business memory machine-readable, so AI can reason with context, not just retrieve data.
Why do AI agents need computable business meaning?
Most organisations already have some of the building blocks. A glossary adds definitions. A taxonomy adds structure. An ontology adds explicit relationships, rules and constraints. A knowledge graph connects that model of meaning to actual data.
List ➔ Glossary ➔ Taxonomy ➔ Ontology ➔ Knowledge Graph
The point is not to create more data. It is to make the meaning already inside the organisation usable by AI.
What does Swarm-ready data look like?
For our second session at CDOIQ, we put theory aside. Matt McQueen, Principal Consultant at Ortecha, demonstrated how a semantic data product could be built and queried using AI.
Using a banking example, the live demo showed you can connect clients, products, customer information, marketing consent and eligibility criteria through an ontology and knowledge graph. When asked which customers should be recommended a particular savings product, the system could use the encoded facts, relationships and eligibility rules to identify suitable customers and explain why.
The generative AI provided the natural-language interface and flexibility. The knowledge graph provided grounded facts and rules. Together, they demonstrated a neuro-symbolic approach: combining probabilistic generative AI with deterministic, rule-based reasoning.
The aim is not to remove generative AI’s creativity. It is to ground it.
Watch the demo:
Sometimes, 'I don't know' is the right AI answer
One of the main points we wanted to demonstrate was that not every query returns and answer. And in enterprise AI, that can be a success.
If an agent does not have the evidence required to make a decision, recognising that limitation can be more valuable than generating something plausible. An unsupported answer from one assistant is a risk; an unsupported decision passed from agent to agent and acted upon across a process is a bigger problem.
“If AI is telling me you don't have the data to answer this question, that's a lot better than it creating a response based on fictitious or internet-based data.”
The goal is trusted execution
Swarm Intelligence is where we believe AI could go next. Connected agents could help organisations execute faster, remove repetitive work and give people more space to focus on judgement, creativity and critical thinking.
But the number of agents is not what will make that vision intelligent. The foundations will.
Context. Agents need to understand what information means inside the organisation.
Governance. They need clear rules around the data, decisions and actions they can access or execute.
Evidence. Decisions need to be traceable back to underlying facts.
People. Human roles need to be deliberately designed around judgement and accountability.
This goes beyond just AI technology. It is a data challenge, a knowledge challenge, a governance challenge and a people challenge.
The organisations that solve those challenges will not simply have more AI agents. They will have AI that understands how their business works.
Make your data Swarm-ready
AI agents are only as useful as the context they can act on. If you are exploring agentic AI, now is the time to ask whether your data and knowledge foundations are ready for it.
Ortecha helps organisations connect data, technology, business meaning and governance so AI can move beyond answering towards trusted execution.
Let's do a quick recap
What is Swarm Intelligence in AI?
In enterprise AI, Swarm Intelligence means multiple AI agents working together across connected tasks and processes. Instead of acting as isolated assistants, they can share context, coordinate decisions and take action using enterprise data.
How is Swarm Intelligence different from agentic AI?
Agentic AI usually describes a single AI agent that can pursue a goal, make decisions and take action with some autonomy.
Swarm Intelligence takes this a step further. Multiple agents work together, which means they need to coordinate, share context and operate within consistent rules. Governance must cover the whole process—not just the actions of one agent.
Why do AI agents need knowledge graphs and ontologies?
AI agents need more than access to data and documents. They need to understand what that information means in the context of the business.
Ontologies define the concepts, relationships, rules and constraints that matter.
Knowledge graphs connect those definitions to real data.
Together, they give agents a clearer and more traceable foundation for reasoning, rather than asking them to rely on statistical patterns alone.
What does it mean to make enterprise data Swarm-ready?
Swarm-ready data is data that agents can understand and use safely. It reflects how the business actually works, with definitions, relationships, rules and context made explicit and machine-readable.
It also sets clear boundaries: what agents can access, which decisions they can make, what actions they can take and when a person needs to step in.
Stephen Gatchell
Partner & Head of AI Strategy, Ortecha
Matt McQueen
Principal Consultant, Ortecha
