Bombardier's sales agent does a 4-hour range calculation in 20 seconds

Bombardier's sales team now gets aircraft range calculations from an AI agent in 10 to 20 seconds. The same calculation took its sales engineers at least four hours, according to [a customer story Salesforce published on 6 November 2026](https://www.salesforce.com/customer-stories/bombardier/agentic-sales-engineering/). Salesforce sells the software the agent is built on.

The job is answering a buyer's technical question during a sale: can this jet fly San Francisco to Tokyo nonstop tonight with five passengers? The answer depends on weight, runway, wind and weather. Until now it came from what the story calls a small, highly specialised sales engineering team, and sellers outside its time zone waited up to four days for answers and revisions.

What the agent does

• Sellers ask the agent, named Clara, in a chat window inside Sales Cloud, Bombardier's customer database, from a phone, tablet or computer.
• For a range question, Clara turns the request into inputs and checks speed, payload and weight. The arithmetic is not done by the AI model: it is done by existing calculation services from Aircraft Performance Group (runway limits and maximum take-off weight) and Bombardier (seasonal winds). The result is drawn as a map, and every assumption and revision is saved.
• For specification questions and comparisons with competitors' aircraft, Clara searches more than 500 technical documents and answers with citations to approved sources.
• Sellers can ask Clara to turn an analysis into a customer email.

When it comes to the computations we've built into this agent, sellers can now generate results on the fly, within seconds.

Fauve Castagna, Manager, Sales AI & Enablement, Bombardier, in Salesforce's customer story

Salesforce reports a 97% accuracy rate on complex technical calculations, a user satisfaction score of 5 out of 5 at launch, and a build of 15 weeks. Bombardier is extending Clara to more than 300 sales staff worldwide.

The people side

The story does not say how many sales engineers Bombardier has, what they do now that sellers ask the agent first, or whether any role changed. It does not say who reviews a calculation before a seller shows it to a buyer. It says validated answers help sellers respond "without overcommitting on aircraft capabilities", which is the risk if an answer is wrong.

The catch

The 97% comes with no method: not what it was checked against, on how many calculations, or by whom. If it means 97 calculations in 100 are right, three are not, and the story does not say how those are caught. The four-hour baseline is an estimate of engineering effort, with no count of requests a month. The satisfaction score has no number of respondents. There is no win rate, deal length or revenue figure.

No cost is given. Salesforce's own engineers and its Professional Services arm did the initial build, and Bombardier pays for Signature Success, Salesforce's premium support plan, whose team helped "close licensing gaps" and fix problems after launch. The result also depends on something most sales teams lack: the calculations already existed as software the agent could call.

What to do if you run sales or revenue operations

• Count the technical questions that leave your sellers for a specialist each month and how long each waits. That count, which Salesforce's story does not give for Bombardier, is the business case.
• List which of those answers come from a documented calculation or an approved document, and which come from one person's judgement. Only the first kind is a candidate for an agent.
• Before a pilot, get the vendor's accuracy figure defined in writing, and name the specialist who reviews answers before they reach a buyer.