Automation in action: Lessons from PMC Treasury’s AI journey 

How can mid-market businesses adopt AI successfully? PMC Treasury’s experience shows the importance of starting with a practical use case, strong data foundations and clear view of how technology should support, not replace, expertise.

The advance of AI has been rapid – yet uptake among mid-market businesses has been slower and more uneven. This is down to a myriad of reasons: What will it cost? Where do I start? The success of businesses that have embarked on the journey show it’s worthwhile, but the path is uncertain without expert guidance.

Until recently, AI had not been a major part of the technology story at PMC Treasury, a specialist advisory firm helping companies manage treasury and market risk. “We’re a people business essentially, but we’re also tech-enabled,” says Tom Baker, Chief Transformation Officer at PMC. “Up until about a year ago, AI hadn’t been a major focus for us,” he admits

Instead of attempting a wholesale transformation, PMC began by defining what AI should mean for the business. Five principles emerged: start small and prove value; augment intellect rather than replace human judgement; commit real resource; choose use cases for learning as well as impact; and treat data quality as a prerequisite, not an afterthought.

Identifying use cases with AI

PMC set out to identify pain points across key use cases, mapping end-to-end process in detail across business development, client delivery, central functions and leadership. Three areas stood out: a sales opportunity research agent, a debt management tool and a business development effectiveness tool.

The first went live a month ago. ‘ORA’ – the Opportunity Research Agent – addresses a simple but significant problem: time. PMC’s consultants were time-poor, expected to research live opportunities, understand the client context, review previous interactions, devise an outreach strategy and act quickly. The pace and volume meant the quality was inconsistent.  “At times our consultants were rushing the process, approaching it inconsistently, or not always doing it as well as they could,” Tom recalls.

ORA changed that by pushing an automated research pack to consultants each morning. For every live opportunity in their pipeline, the pack includes a deal synopsis drawing on live news and deal databases, a portfolio company profile, historical CRM activity, previous wins, recent interactions and key contacts.

Crucially, the consultant remains an important part of the process. “The goal is to make it as easy as possible for them,” says Rory Cooke, Senior Data Analyst at Inflexion, which has backed PMC since 2018. One-click buttons generate pre-populated outreach emails for consultants to review and tailor before sending. “You clearly could have just taken the person out of the loop completely,” Rory says, “but we want the consultant to remain involved as the final authority.”

The ‘one-click solution’ wasn’t the original plan. Consultants assumed they wanted a chatbot, but testing told a different story. “The email pushed straight to the consultant’s inbox got more traction than the chatbot, so pushing the output to them was the right route,” says Tom.

Automating with AI

The build itself consisted of three components: data, intelligence and automation. PMC’s existing Snowflake datalake gave the project a strong foundation, bringing together CRM, relationship and business data into one place. A large language model provided the intelligence layer, structuring information and turning it into usable insight. A workflow automation tool then acted as the glue, moving information between systems and triggering the daily output.

The technology was surprisingly simple to work with, according to Rory. “It’s not writing code; it’s essentially sticking together a workflow diagram. If you’ve got somebody who is a competent Excel user, these tools are really pretty easy to get to grips with.”

The simplicity belies the prep work required. The datalake had taken six months to build and was designed for reporting and analytics; it was not created specifically for AI. “We were spending too long in meetings debating the numbers,” says Tom. “People would turn up with different interpretations of the data, each backed by their own charts and narrative.” That debate has now largely disappeared. “The truth is on our datalake – there is no debate about whether that data is right or wrong,” Tom explains. AI readiness became a second-order benefit.

How Inflexion supported PMC’s AI adoption

Inflexion worked closely with PMC through the early design and build, helping the team get from idea to prototype in a matter of days once the data foundations were in place. “The skill is not just deep technical expertise, but knowing what to build,” Rory says, adding that it was ultimately an analyst at PMC who proved instrumental to the build and ultimately success – bringing internal knowledge and commercial acumen to the process. somebody who is commercially minded and knows the business. 

Inflexion’s support helped turn PMC’s intent into action. "We had the datalake in place but were struggling to find the right direction. It was Inflexion's in-house team that really helped us get moving," says Tom. 

The early results are encouraging. After four weeks live, average time to first outreach fell from around 4.4 days to 3.3 days, with fewer opportunities receiving no outreach at all. PMC has also seen early signs of more opportunities progressing to proposal stage, alongside four new introductions to companies it had not previously engaged.

Perhaps most tellingly, one of the most junior members of the team, less than a year out of university, has become one of PMC’s most prolific organisers of client meetings. “It’s quite exciting to see that someone can come in with some thoughtful use of the tool and actually make a real impact on the business,” says Tom.

The success marks the start of a broader programme at PMC , with three more use cases under development, and the business hiring for a dedicated AI role.