The tools to automate rule-based work have existed for ten years. Agents now handle the parts that were too variable to automate. We build workflows and agents on Microsoft Power Platform, in your tenant, and have done so for six years.
For years the barrier to automating a process was that every path it could take had to be written down first, and most processes have too many. The predictable work was automated. The rest stayed with people, often a team working from a shared mailbox.
Agents remove that barrier. An agent can read what arrives, work out what it is and decide what to do with it, without every case being scripted in advance. Not every step should be an agent. Most of a process is ordinary fixed steps, and a workflow should run those because it is cheaper and predictable. The agent takes the steps that need judgement, and a person handles the exceptions. Copilot Studio has recently been rebuilt around this combination of workflows and agents.
The harder part is knowing where the value is. Productivity uses are easy to see, which is why most agents so far help individuals write and search faster. Process uses are harder to see, because few people in an organisation see a whole process end to end. Every engagement starts with an analyst working alongside the team that does the work, before anything is built.
Three weeks to establish whether a use case can be built and whether it is worth building, with the answer in writing.
Our team works directly with your business from week one on the outcome you want. Within weeks a process, app, agent or site is live in your tenant. Fixed price.
The larger builds: portals, integrations and case systems, scoped and priced before work starts.
A monthly retainer for the AI council, the governance baseline and training, with people who already know your business.
Most clients take them in this order. Any of the four can be the starting point. How each one works
Trusted by
What we brought back from Microsoft's Copilot Studio bootcamp in London, in nine ideas: three trends in how agents are used, the technology that meets them, and three things to do about it.
The tools to automate a whole process properly have arrived, but the rush to AI skipped the old discipline. Analyse, eliminate, simplify, then automate, worked through one law firm matter.
A lawyer hands four contract checks to her Copilot and saves fifteen minutes a contract. She was right to, but the matter takes as long as ever, and that gap is what process automation is for.
The first conversation is thirty minutes with Tony Brooks, no deck. Describe the job and you will get a straight answer on which of the four fits, or that none does yet.