By Sagi Eliyahu
Excitement about the transformational potential of AI in the enterprise has never been higher.
It’s manifested most recently into the explosion of interest in AI agents. Agents are specialized, AI-powered software programs that can work autonomously to achieve long-term goals.
As Meta’s head of business AI, Clara Shih, recently told CNBC, Meta expects “every business” to use AI agents “the way that businesses today have websites and email addresses.”
The hype around agents is justified, but discussions about how the enterprise should work to achieve Meta’s vision remain incomplete. To get transformational value out of agents, it matters greatly not just how powerful or plentiful they are, but how strategically and thoroughly you integrate them into the infrastructure of your day-to-day operations.
For this you need a means. The only adequate means currently available is orchestration.
Orchestration is technology that undergirds the whole of your organization’s operational infrastructure, including your tech stack, your people and your policies. Organizations use orchestration to connect and automate processes across these various moving parts — allowing you to bridge the gaps between human collaboration and technology, and put structure around intelligent entities.
That’s critical. Intelligent entities — all intelligent entities — need structure to work effectively. This is as true of humans and human-led organizations as it is of AI. You want intelligent entities to be able to work autonomously and creatively in pursuit of the goals you set for them. But to effectively pursue those goals, you also need direction and hierarchy, governance and org charts, processes and rules.
Without this, we work inefficiently in the best case and destructively in the worst. Even genius workers, without governance, mostly produce only chaos. You can have a great violinist. You can put them next to a pianist and a singer and you have a band, and you don’t need orchestration for that. But if you have 130 musicians and you put them in a room, you need a way to get them to play together. Otherwise it’s literally just noise.
It was for this reason — to harness and manage intelligence, determining what it does and, importantly, what it should not do — that mankind invented organizational processes and policies in the first place.
The importance of putting structure around intelligence and skill only becomes more pronounced as the intelligence in question becomes smarter and more capable.
Orchestration helps us do so in the context of AI. Here are just a few examples of what that looks like.
Connectivity
The first has to do with connectivity. To provide transformational value, AI agents must be able to autonomously conduct work across all your organization’s tools, systems and teams. That means you need a way of connecting all your tools, systems and teams. Agents that help you navigate just one walled user interface don’t do much to move the operational needle. Orchestration platforms help you do this, so your agents can navigate your tech stack seamlessly behind the scenes.
Accessibility
Another has to do with accessibility. For AI agents to be useful, what you need are agents that are intelligent and specialized and that, importantly, process designers can make available to employees when and where they need them — in the environments where they already work, and in accordance with how they like to work. Orchestration allows you to create and deploy agent-powered processes that do precisely that.
For example, an agentic orchestration platform that comes with an agent builder can help internal teams create and orchestrate processes that automatically surface AI agents for employees when and where they need it.
Employees can also easily access AI agents using an AI front door inside Slack, Teams or wherever else they like to work. They ask their question, and an agentic orchestration platform will tap whichever agent is most appropriate, depending on the request.
That agent — or whichever other agents that agent decides it needs to collaborate with to give the employee what they need — will guide them to resolution. Orchestration handles all the toggling between systems that employees used to have to do manually in the background.
Governance
Another example has to do with governance. Orchestration platforms give organizations a behind-the-scenes means of determining what their agents can and cannot do, as well as what they can and cannot access. In this way, you can ensure that an LLM never accesses login credentials, handles sensitive data, or takes certain actions that enterprise organizations would not trust an LLM with.
Anywhere you wouldn’t want pure autonomous automation, you can put regimented processes in place.
So, too, can you define when in your various internal processes agents should tap humans for help. (Humans are always kept “in the loop.”)
Orchestration has not commanded the same attention in the marketplace as the race to build powerful reasoning models or the rise of agents themselves. But that will change as the backbone necessity of orchestration gets borne out.
Organizations that use AI to win the future won’t be those with the “smartest” AI, but those using AI the smartest. That means putting structure around intelligence. That requires orchestration.
Sagi Eliyahu is the co-founder and CEO of Tonkean, an AI-powered intake and orchestration platform that helps enterprise shared service teams such as procurement, legal, IT and HR create processes that people actually follow. Tonkean’s agents use AI to anticipate employees’ needs and guide them through their requests.
Illustration: Dom Guzman
Stay up to date with recent funding rounds, acquisitions, and more with the
Crunchbase Daily.