Mott MacDonald staff have created more than 8,000 AI agents via Microsoft Copilot. The agents support everything from project assurance and stakeholder engagement to design automation and data-driven decision-making.
In an interview with Microsoft, MottMac global head of AI and data, Nasrine Tomasi, detailed the staggered roll-out of Copilot across the business and its impact. MottMac started with 50 Copilot licences last year, expanding to 1,000 and then 5,000 licences. Half of the new licences in the second stage went to early-career professionals, many of whom were already comfortable using generative AI.
Examples of the agents include one on an airport project that helped MottMac work through around 220,000 requirements and flag about 200 conflicts for attention. On a large underground station, structural designers used Copilot-supported analysis to show that temporary works could safely form part of the permanent structure, reducing the amount of additional construction required.
AI agents have also been deployed in contract bidding: a hydropower team has built a collection of agents to help prepare business proposals for new projects. They assemble requirements, calculate budgets, and find matches to the company’s capabilities. Team members refine the agents after each bid and have set a target to cut the average proposal process from one week to 48 hours.
“They are all contributing to the agents to make them better with every bid,” Tomasi told Microsoft. “And they are winning more work as a result. We had talked a lot about productivity and doing things faster, especially a year ago. But we are now going way beyond this, driving up quality and creating value for our clients.”
Meeting the challenges of such extreme growth
Such extreme growth is not without its challenges: some overlapping solutions, dormant agents, and knowledge gaps when creators leave or change roles, for example.
To tackle this, MottMac is creating a central AI library, so if someone is planning to develop an agent for proposal writing, for example, they’ll be able to discover whether one already exists that they can use or adapt. The aim is not only to improve visibility, but also to identify successful innovations with the potential to scale across the organisation.
Meanwhile, MottMac’s original AI assistant, EMMA (Every Mott MacDonald Answer), is now moving beyond knowledge retrieval and being deployed in business processes.
Monthly project reviews are one example. Project managers and project principals previously had to complete a standard form covering the health and status of a project. Now, using an app hosted in Microsoft Teams, EMMA can capture a structured conversation between them and organise the notes around an agreed agenda, highlighting lessons learned and future actions. More than 10,000 lessons have been captured since the app launched in May.
“The richer the conversation, the richer the notes that you’re having at the end of this process,” Tomasi said.
MottMac is now rolling out Copilot to all 20,000 staff. The business is also preparing to use Microsoft Fabric as a central data platform, giving Copilot and its agents better access to the information they need.
