What Walmart's chatbot means for your commercial organisation

In early 2021, retail giant Walmart had a hundred suppliers negotiate with a chatbot. These were the long tail of its supplier base: suppliers of shopping trolleys or fleet services. Walmart's buyers barely had time to talk to all of them. Hiring extra buyers would have cost more than it brought in. So those contracts rolled over for years on standard terms.

The chatbot could negotiate with every one of those suppliers, all conversations at once. Walmart expected a new agreement with one supplier in five. In the end, it reached one with almost two in three. Walmart saved 1.5% on average and extended payment terms to an average of 35 days, according to Van Hoek and colleagues (2022). After that success, Walmart extended the approach from Canada to the United States, Chile and South Africa.

The thinking done up front

The success came mainly from what happened before the chatbot sent a single message. Before the first negotiation, Walmart decided on four things:

• Which suppliers take part. Only suppliers of goods for Walmart's own use, not of products on the shelves.

• Which goals it wants to reach. A lower price or a longer payment term.

• The room to negotiate. A longer notice period for the supplier, or more volume in exchange for a discount.

• Who approves those limits. Alongside five buyers, people from finance and legal took part in the pilot.

The buyers also gave the chatbot scenarios of successful negotiations. Only because the rules and the context were ready could the bot hold hundreds of conversations at once, without a buyer stepping in.

On the other side of the table

Now put yourself in the supplier's position. What would this mean for you? One renegotiation would be no problem. Your managing director or your inside sales team sorts that out in an afternoon. But imagine every large customer using a chatbot like this. You would get requests for new prices or terms every week, from all those customers at once. For each request, someone in the company has to decide how far you can give way.

Inside your own company, too, AI agents, bots and automated processes let your team do far more in far less time. Your digital marketer writes a quarter's mailings in a week. Your salesperson builds his own configurator with the prices for his customers. Still, not everything can simply go out to the customer. A mailing with a new promise, or a price below the agreed margin, still needs a decision from someone with the mandate.

Where your organisation gets stuck

With your customers and in your own team, it comes down to the same thing. AI produces far more price requests and mailings than before. Most of them need no decision. But as the total grows, so does the number of exceptions that do, such as an extra discount or a new promise. Those decisions become the bottleneck of your commercial organisation.

I redesign commercial organisations every day, and I build, write and work with AI myself. With my clients, I already see that bottleneck forming. I wanted to know whether it also happens beyond my own clients, so I read a good deal of research, mostly on the time AI saves and on reviewing what AI produces. From those studies and my own practice comes this proposition for your commercial organisation:

AI makes creation cheap. Checking and deciding become the new bottleneck of your commercial organisation. Redesign your organisation with rules and context so that checking runs more smoothly.

From a single task to the whole organisation

Most research on AI measures single tasks: how fast a quote is ready, how many errors a text contains. I found no research on the impact of AI on an entire commercial organisation. So it seemed worthwhile to put the research on AI next to my own framework for designing commercial organisations. That framework describes how a company works commercially, from the commercial policy down to the daily work of each employee. I developed it to see quickly, with my clients, where their organisation gets stuck and what needs redesigning.

In the next parts, I work out the proposition step by step:

1. The framework. A commercial organisation in three levels and six steps, with gates in between.

2. More is being made than ever. AI makes creation cheap, in your own team and at your customers. The volume on your desk grows with their use of AI.

3. Checking becomes the bottleneck. Reviewing takes more time than creating. A fluent AI text hides its errors.

4. Judgement can be built in. With written rules and context, the ordinary cases go through smoothly. Only the doubtful cases still reach a person.

5. Every role changes. Roles move from making things towards more human contact and towards guarding the rules.

6. Write down your rules. Where to start today, before your team uses AI widely.

Walmart made its decisions up front, before the chatbot started. You will find decisions like these in every commercial organisation, but they are rarely written down. Where exactly they sit is the subject of the next part, on the three levels and six steps of your commercial organisation.