InvestNot yet confirmed elsewhere1 publisher2 min readPublished
AWS puts a chatbot in front of its 18-tab deal-pricing spreadsheet
AWS's pricing team turned an 18-tab Excel model for evaluating customer deals into a chatbot, finance chief John Felton told Fortune. It gives pricing staff a plain-language way to question that model inside a $169 billion annualized business.
The Investor · Invest desk
What happened
- Pricing staff can ask the chatbot what happens to a deal if the price falls 20%, how different payment terms change it, or where its break-even point sits.
- A second finance team built an AI agent that checks payment terms in every customer contract against AWS's payment system, where staff used to check only a sample.
- Felton's instruction to AWS finance staff is to "Use AI every day", and he tracks whether they follow through.
- Felton said customer AI talks two years ago centred on productivity and cost cuts and now turn more on new revenue products and customer experiences.
Why it matters
- capability AWS's payment-terms control now covers the whole contract population, so a mismatched term can no longer slip through by falling outside the sample.
- constraint If many customers must finish cloud and IT modernization before deploying AI at scale, their AI spending with AWS has to wait for that migration work.
- decision With no central list of tasks or tools, AWS finance is not standardising on one AI product, and its adoption is judged on internal metrics Felton has not detailed.
"Replaced" overstates what Fortune describes. Staff who prefer the spreadsheet can still export the model to Excel [4]. The 18 tabs still exist, and the chatbot is a new way of putting questions to them in plain language [1].
The case for calling it more than a front end comes from Fortune itself. Sheryl Estrada wrote: "The point isn't simply that AI made an existing tool easier to use. It changed how employees interact with the underlying analysis." [12] For a deal model, that matters. On the annualized revenue figure Fortune cites, one percentage point of price, held across the whole book, would be about $1.69 billion a year [14]. I'd expect a model that can be questioned in a sentence to be consulted on more deals, and by people who never learned the tabs.
From here the tool can go one of three ways. It can stay a front end on a model people maintain by hand, with the gain showing up as analyst time and never in reported margins. It can spread past the pricing team, and then the control question is who checks an answer generated from a typed sentence before it shapes a customer offer. Or the pricing logic can migrate out of the spreadsheet into the AI system. Only the third puts AI in charge of the commercial call.
The first path fits the evidence best. The export survives, and every example query Fortune gives concerns a single deal [3]. A disclosure that the chatbot recommends or approves prices, or that the spreadsheet has been retired, would prove that view wrong.
The contract agent sits closer to cash. When the payment term recorded in the system differs from the one in the signed contract, it changes when the money arrives, or whether it does. "Let's not just do a sample size, let's do all of it," Felton said, describing the team's approach [6].
Felton, AWS finance chief since 2024 [2], is also CFO of the business that sells cloud and AI to the companies he describes [11]. "I think everyone is still early," he said [8]. He counted AWS among the companies getting value from rethinking a process from the start: "That is where a bunch of customers, including ourselves, really see a lot more value," he said [13].
What to watch
- Whether Felton or Amazon publishes the follow-through metrics behind "Use AI every day", the only adoption measure for AWS finance he mentioned.
- Any count of the contracts the payment-terms agent checks, or of the mismatches it finds; either would put a size on the cash involved.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence35
- Adoption25
- Hype gap+10
- Incentives70
- Confidence40
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
AWS's pricing team had relied on a complicated Excel model spanning 18 tabs to evaluate customer deals, and converted it into a chatbot interface that lets users explore scenarios through natural-language questions.
- [2]
John Felton is SVP and CFO of AWS. He has been with Amazon for more than 20 years, previously led Amazon's worldwide operations, became AWS finance chief in 2024 and reports to Amazon CFO Brian Olsavsky.
- [3]
Users of the pricing chatbot can ask what happens if a price falls 20%, how different payment terms affect a deal, or where the break-even point lies.
- [4]
Employees who prefer the spreadsheet can still export the model to Excel.
- [5]
Another AWS finance team built an AI agent to compare payment terms in customer contracts with those recorded in AWS's payment system. Previously employees checked a sample of contracts; the agent lets the team check all of them.
- [6]
"Let's not just do a sample size, let's do all of it," Felton said, describing the contract-checking team's approach.
- [7]
Felton's instruction to AWS finance employees is "Use AI every day." He does not dictate which tasks or tools they use, and tracks whether employees follow through, though he did not detail those metrics.
- [8]
"I think in general, I would say it feels early across the board ... I think everyone is still early," Felton said of AI adoption among companies.
- [9]
Felton said customer AI discussions about two years ago focused largely on productivity and cutting costs, and now: "There really is a lot more focus on: how do I use AI for new revenue products, for new customer experiences?"
- [10]
Many AWS customers are still early in their broader cloud and IT modernization efforts, a prerequisite for deploying AI at scale.
- [11]
AWS, Amazon's cloud computing and AI platform, is now a $169 billion annualized business.
- [12]
"The point isn't simply that AI made an existing tool easier to use. It changed how employees interact with the underlying analysis."
- [13]
"That is where a bunch of customers, including ourselves, really see a lot more value," Felton said, referring to rethinking processes with an AI-first mindset.
- [14]
One percentage point of price, if it held across AWS's whole $169 billion annualized business, would be about $1.69 billion a year.
Sources
1 independent publisher whose own reporting we read for this story.
- fortune.comAWS pricing team turns 18-tab Excel model into an AI chatbot to evaluate customer deals, CFO says
1 article · October 8, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Entities
- Amazon Web ServicesFollow
- AmazonFollow
- John FeltonFollow
- Brian OlsavskyFollow
- Amazon QuickFollow
- Microsoft ExcelFollow