The challenge

Our client manufactures industrial equipment and complete production systems.

The individual machines and components are largely standard. What is rarely standard is the customer’s final requirement.

A customer may need a different capacity, a modified configuration, alternative utilities, equipment added or removed, or changes dictated by an existing facility.

So while every quotation starts from something the company has sold before, almost every one requires some engineering judgement.

That makes quotation creation much more than filling out a sales document.

It requires reconstructing the closest configuration the company has delivered before, understanding what changed, and determining how those changes affect the rest of the system.

Where the time was going

The sales team consisted of four people.

A typical quotation could take roughly a day to prepare.

Very little of that time was spent actually writing the quote.

The real work was finding the right precedent.

The company had more than a decade of historical quotations across its ERP system, spreadsheets and other documents. In theory, this represented an extremely valuable body of knowledge.

In practice, finding the right historical quotation depended heavily on memory.

An experienced salesperson might remember that a similar configuration had been quoted several years earlier and know approximately where to look.

A newer salesperson usually could not.

When the sales team could not find the answer, they turned to engineering.

As a result, preparing quotations created repeated interruptions for engineers who were being asked to reconstruct information that often already existed somewhere in the company’s history.

What we built

We built an AI-assisted quotation workflow around the company’s existing quotation history.

Given a new customer requirement, the system finds genuinely comparable historical quotations and uses them to assemble a proposed equipment and item list.

Instead of starting with a blank document, the salesperson starts with a grounded draft.

They can then:

  • change capacities
  • add or remove equipment
  • replace components
  • correct assumptions
  • adjust specifications
  • apply commercial judgement to pricing and payment terms

The salesperson remains responsible for the final quotation.

The objective was not to automate commercial or engineering judgement. It was to eliminate the repetitive reconstruction work surrounding it.

Making the historical data useful

The key requirement was that the system should not invent a plausible-looking quotation.

Suggested items and configurations are connected back to actual historical quotations, allowing the salesperson to see the precedent behind a recommendation.

That makes suggestions easier to verify, adjust and trust.

For a system like this, trust does not come from the AI always being right.

It comes from making its recommendations checkable.

The outcome

Average quotation preparation time fell from approximately one working day to around two hours.

But the impact went beyond quotation turnaround.

Salespeople became less dependent on engineering to locate previous configurations and answer routine questions.

Engineering therefore spent less time acting as an internal search function and could focus more of its attention on work that genuinely required engineering judgement.

The company’s historical quotation archive also became useful to a much wider part of the organisation.

Previously, much of its value was locked inside the memories of experienced employees.

Once that history became searchable, someone who joined recently could benefit from work the company had done years earlier without first knowing that a particular quotation existed.

The broader lesson

This problem is common in businesses that sell standard products configured into customised solutions.

The apparent bottleneck is often engineering capacity.

But a significant amount of the delay can come from something simpler: the organisation has already solved a similar problem before, but nobody can easily find the answer.

Years of quotations contain a record of:

  • customer variations
  • configuration decisions
  • substitutions
  • equipment combinations
  • exceptions to the standard catalogue

The catalogue describes what a company sells.

Its quotation history describes how those products actually get adapted to customer needs.

Making that history searchable can therefore create value long before a company tries to automate the entire quotation process.

How Nond helped

Nond helps businesses turn fragmented operational knowledge into AI systems that make existing teams faster without removing human judgement.

If complex quotations still require people to search through years of old files or ask around for the closest precedent, we’d be interested in hearing how the process works at your organisation.