Commercial and merchandising is one of the most decision-dense functions in any brand business and one still largely run on a weekly clock. This is where AI stops being a faster report and starts changing how the decisions get made.

Where the work is decision-dense

If operations are where AI proves itself first in a physical business, commercial and merchandising is where it changes the most valuable decisions. Deciding what to make and buy, how to price it, how to promote it, and how to sell it into accounts and channels is dense with judgement, and it runs on a constant stream of data: sell-through, stock cover, competitor moves, weather, season, margin. That combination - high-stakes decisions, made repeatedly, on lots of data - is precisely what agentic AI is suited to support. It is why advisers watching the sector consistently point to assortment, pricing and promotions as the places the first practical changes show up.

The pattern today: decisions on a weekly clock

In most commercial teams the rhythm is a cycle. Reports are compiled, often by hand, pulling numbers from several systems into a spreadsheet. The team meets. Decisions are made against last week’s picture. Then the cycle repeats. A merchandiser or account manager can spend more of the week assembling the view than deciding anything with it.

The problem is that the market does not run on a weekly clock. A line sells out in one region while sitting untouched in another. A competitor drops price mid-season. A promotion underperforms in its first days, with no easy way to react until the next cycle. By the time the report is built and the meeting has happened, the moment to act has often passed. The season moves faster than the cycle that is meant to manage it.

Three decisions AI changes first

Assortment and range. Agents watch performance continuously across brands, categories and channels, and surface where a range is over- or under-invested - what to chase, what to cancel, where demand is shifting - with the numbers already assembled rather than compiled after the fact.

Pricing and markdown. This is where the shift is sharpest. Instead of a fixed markdown schedule, a pricing agent can weigh competitor pricing, current stock position, weeks left in the season and margin target together, and recommend - or, within limits, set - the price that best balances them. For a seasonal brand business, getting markdown timing right across a range is worth a great deal of margin.

Promotions and trade investment. Agents track how a promotion or trade plan is performing against expectation in near real time, and flag where spend should be rebalanced across accounts, channels or categories before the money is wasted, rather than after.

The role shifts from assembling to deciding

The common thread is that the work moves up. When the assembling is done - the reports built, the options ranked, the impact on sales, margin and working capital already calculated - the commercial person is left with the part that needs them: the judgement. Give every buyer the equivalent of a team of analysts working continuously, and the job stops being about producing the numbers and starts being about deciding what to do with them.

“The role stops being about assembling the numbers and starts being about deciding what to do with them.”

Two things make this practical rather than aspirational. The recommendations come ranked and explainable - tied to the sales, profit and working-capital impact - so a person can trust and interrogate them rather than take them on faith. And increasingly the interface is conversational: a commercial manager can put a plain-language question to the plan and get a plain-language answer, instead of navigating another dashboard. The same intelligence sharpens the moments that have always mattered most, such as walking into an account or supplier negotiation already knowing exactly how each line and category is performing.

Crawl, walk, run

It is worth being honest about sequence, because the failure mode here is jumping straight to autonomous pricing and frightening everyone. The credible path, as advisers such as Bain describe it, is staged. First, agents assist - consolidating data and surfacing insight. Then they recommend - proposing assortment, pricing or promotion actions for a person to approve. Only then, and only where it is safe, are they wired into execution to make changes directly, always under human oversight and within agreed guardrails. You earn autonomy decision by decision, as trust and evidence build. Starting with recommendation, not execution, is what keeps the team on side.

On the systems you already run

As with operations, none of this means replacing your commerce or planning systems. The redesign connects to the systems that already hold your product, pricing, sales and customer data. SAP and Google Cloud both now provide managed ways to build and orchestrate these agents, and both support open protocols - including Model Context Protocol - that let agents read from and act on your existing estate. Specific capabilities are at different stages of release, so confirm the detail at the time of reading. The work is in redesigning the decision, not migrating the platform.

Where the value shows up

The value here is margin, not headcount. Better assortment means less capital tied up in the wrong stock. Better markdown timing protects margin that is otherwise given away. Better-targeted promotion means less wasted trade investment. And the merchandising and commercial teams spend their time on the decisions that move the number rather than on the reporting that describes it. Adoption is still early - industry surveys in 2026 suggest only a minority of large organisations run these workflows at scale, though most are piloting or planning to - which means the advantage still belongs to those who move deliberately now.

Start with one category or account

You do not rewire the whole commercial function at once. Pick one category, one channel or one account where decisions are visibly slowed by the weekly cycle. Map how a decision actually gets made today - where the data comes from, who assembles it, how long it takes - and redesign it so agents do the assembling and the monitoring while your people make the calls. Prove it there, keep the human firmly in the decision, and the pattern extends across the function.

In this series we have argued for redesigning work rather than speeding it up, shown what a parallel workflow looks like, and walked through operations. Commercial and merchandising is where those ideas meet your margin. If you would like to see one of your own commercial decisions redesigned this way, that is the kind of session we would start with. Get in touch now.