By Leigh Walters, CEO, TRG Screen
In 1913, Henry Ford didn't build a better car. He built a better way of building cars. Model T production fell from 12 hours to 90 minutes. The price dropped from $825 to $260. Car ownership stopped being a luxury and became normal.
A recent McKinsey article, The AI assembly line: Strategic imperatives for CEOs, argues that agentic AI is doing to cognitive work what Ford did to physical labor. I think they're right.
More than a century later, the same principle continues to appear in modern industries. SpaceX has accelerated rocket development not simply through better engineering, but by shortening the feedback loop between design, manufacturing and testing. Elon Musk often describes this as building "the machine that builds the machine", creating systems that improve the speed at which organizations learn and act.
Ford's assembly line may be one of the earliest examples of this thinking. SpaceX is one of the most modern. Agentic AI may be the next.
And I think managing market data inside a financial institution is one of the clearest examples of where it has to happen next.
Today, much of that work remains highly manual. Invoices arrive in dozens of formats from hundreds of vendors. Contracts hide commercial terms in clauses written years ago by people who have long since moved on. Usage data sits in one system, billing in another, the agreement in a third, and the budget in a fourth. The people who can reconcile all of that are scarce and overworked, and the work piles up faster than they can clear it.
Most market data teams I speak to are running a fairly manual production line, and the production line cannot keep up.
The problem is that people have become the ‘integration layer.’ Highly skilled market data professionals spend a significant amount of time manually keeping information and tasks moving across the process. The work gets done, but it relies heavily on human effort. As complexity grows, so does the effort required to keep the machine turning.
The bigger problem is where effort is spent. Instead of focusing on optimization, commercial outcomes and decision-making, experienced people spend their time operating the process itself.
A year or so ago I shared a view on on LinkedIn that framed this as an assembly line: a standardized, automated, 24x7 conveyor belt running across every step of market data commercial processing, combining robots and humans, with each workstation doing one job well.
The McKinsey article is the strategic version of that picture. When that idea is mapped to your category, and to the jobs our customers tell us they actually need done, they tend to fall along six key workstations, in a fairly linear sequence:
-
Category and cost reporting: Plan and budget what the firm should spend.
-
Order and demand management: Request, approve and provision the services people need.
-
Sourcing and contracting: Negotiate, extract and sign the agreements that govern it all.
-
Inventory management: Keep users, services and entitlements reconciled to reality.
-
Invoice processing: Ingest the invoice, match every line to a contract and to actual use, and resolve what doesn't match.
-
Books and records: Allocate, accrue and report, cleanly, to the people who need to know. Today, these workstations exist in most firms, but they're staffed by people moving data, files and paper between them. In other words, the integration layer is still human.
The promise of an agentic assembly line is that each workstation can be served by a purpose-built agent. One that reads the contract, one that watches the usage, one that processes the invoice, one that spots the anomaly. All sitting on a single shared dataset, all coordinated by an orchestration layer, all running while everyone else is asleep.
The breakthrough isn't automation. The breakthrough is what becomes possible once information can move freely across the lifecycle and the cognitive cost of each task falls close to zero.
McKinsey makes three points that ring loudly for anyone running a market data function:
The real prize isn't automating tasks... it's expanding decision-making capacity. The bottleneck was never that we couldn't process invoices. It was that nobody had time to reconcile the invoice with the contract, with the usage data, with the budget, and with the benchmark, all at once. Agents change that.
The orchestration layer matters more than the individual agents. A clever contract reader and a clever invoice reader sitting in separate apps are two products. The same two agents talking to each other across a shared data foundation are a platform. The platform is where the value compounds, and where the visibility lives.
The lesson is similar to the one Ford and, more recently, SpaceX demonstrated. The breakthrough doesn't come from optimizing individual activities. It comes from improving the flow of tasks or information between them. When every stage informs the next, organizations learn faster, act faster and make better decisions.
This is a leadership job, not a technology project. The hard work isn't just building the models, it's redesigning the workflow, the roles, and the customer promise around what an assembly line makes possible. McKinsey notes that leader-led transformations are 1.5x more likely to succeed than those run primarily by technology teams. In my experience, the multiplier is higher in this category, because so much of the value sits in domain expertise that can only be encoded by the people who already do the work.
Ford didn't win because the Model T was better. He won because he changed the physics of the unit cost. The same window is open now for whoever is willing to rethink how market data work gets done.
Thanks for taking time to read Market Data Matters! Don't forget to subscribe if you haven't yet.
