Skip to content
backendflow.tech

AI assistant

How a tank manufacturer stopped losing requests because of one engineer

Manufacturing of steel tanks and vessels

AI assistant for a steel tank plant
Steel tank manufacturing (vertical and horizontal tanks)
60-150 employees · Volga region
NDA

TLDR

The plant manufactures custom steel tanks: vertical tanks, horizontal tanks, and vessel equipment. Every request requires a technical specification and a commercial offer for a specific facility. Only one person knew how to prepare the specification - the lead design engineer. Sales waited in his queue for 2-3 days. BackendFlow developed an AI assistant: the manager uploads a questionnaire, and the system creates a draft specification and commercial offer in 5-8 minutes.

Result: the designer stopped being the bottleneck. Commercial offer preparation time went from 4 hours to 40 minutes for draft review. The plant processes 30-40% more requests without hiring a second specialist.

Context

A non-standard tank always requires a calculation for a specific facility: medium type, operating pressure, anti-corrosion protection, and installation specifics. Only one person could turn a client questionnaire into a full technical specification - the lead design engineer with years of experience working with standards.

A non-standard request took him 3-5 hours. A standard one took 1-1.5 hours. With 18 requests per month, 40% of them non-standard, about 50 hours per month went only into commercial offer drafts. That is a third of his working time.

For the sales team: response speed to the client depended not on the manager, but on the workload of one specialist. Sending a commercial offer quickly was impossible.

For the designer: a significant part of working time went into preparing drafts from known templates - work he did not have to do personally.

Task

Give the sales team a tool for preparing a commercial offer draft without involving the designer in every request. The designer should review and approve, not create everything from scratch. Final technical responsibility remains with the engineer.

Solution

BackendFlow developed an AI assistant trained on the plant's standard configurations, regulatory documentation (GOST, PB 03-605, OST), and the archive of past projects.

Workflow:

  1. A client request arrives through a questionnaire
  2. The manager uploads the questionnaire into the system
  3. The assistant validates the inputs and requests missing parameters if needed
  4. The system creates a draft: material selection, parameter calculation, and commercial offer using internal pricing rules
  5. The completed draft goes to the designer for review

The designer is involved only for non-standard units and facilities with special conditions. The system is running in the working environment, and before/after statistics are being accumulated.

Insights

Input validation shortened the approval cycle.

A significant part of delays in the old workflow happened before the calculation: the manager sent incomplete data, the designer asked clarifying questions, and the cycle stretched out. Built-in validation removed that loop: the system highlights missing inputs before draft generation starts.

The manager gained a tool to accelerate the deal.

When a commercial offer draft is ready in 40 minutes, the manager can send it to the client on the same day. Response speed became a competitive advantage; previously some clients left for suppliers who answered faster.

The designer's expertise now scales.

Knowledge of standards and standard configurations that previously existed only in one specialist's head has been formalized and made available to the system. If the designer leaves or is temporarily unavailable, the plant does not lose the ability to process requests.

Results

MetricBeforeAfter
Commercial offer draft preparation time3-5 hours~40 min for review
Dependence on the designerCompleteOnly non-standard units
Request processing without hiringLimited by workload+30-40% flow
-85%
designer time spent on a commercial offer draft
$4.5-6.1K
forecast annual TCO (development + support)
15-21 mo.
forecast payback from direct effect

Assumptions: 18 requests/month, 40% non-standard, designer salary about $1.2K/month, adoption 78%. Conservative scenario. Verification after a quarter of system operation.

Want the same kind of calculation for your production?

Book an audit ->
<- All casesNext case ->