Post 19 of 19
Practical uses for AI in ship purchasing, and the controls that keep purchasers in charge of decisions.
AI is most useful in procurement where purchasers spend their time reading, comparing and chasing. It is least useful as an autonomous buyer. At a ship-management company running Delta BI's procurement workflow, an AI layer supports purchasers rather than replacing their decisions. This article sets out where that kind of support helps, and which controls to put around it.
AI works on the data your process produces. If requisitions are free text, quotes live in email and nobody stores price history, an AI assistant can only guess. Structured requisitions, one catalogue and line-level prices come first; AI then makes that data faster to use.
Typical uses that save purchaser time without handing over the decision:
Reading supplier quotations and turning them into lines that can be compared side by side.
Matching free-text requisition lines to catalogue items and IMPA codes, for the purchaser to confirm.
Summarising how quotes differ on price, lead time, delivery port and terms.
Flagging quoted or invoiced prices that are well above the item's history.
Drafting requests for quotation and supplier follow-ups for the purchaser to review and send.
Answering status questions about an order from the data, so the vessel gets an answer faster.
Keep decisions with clear accountability with the people who own them.
Supplier selection and award of the order.
Approval of orders above the agreed thresholds.
Technical acceptance of alternative or non-original parts.
Supplier onboarding and sanctions decisions.
Release of payment.
The aim is that every AI suggestion can be checked, and every decision can be traced.
Show the source document next to every value the AI extracted.
Send low-confidence results to a review queue instead of accepting them silently.
Log each suggestion, who accepted or changed it, and when.
Apply the same access permissions as the rest of the procurement system.
Agree with the model provider how your data is stored and whether it is used for training.
Use a recognised framework, such as the NIST AI Risk Management Framework, and check which obligations of the EU AI Act apply to your use.
Measure AI support the same way you measure the rest of the process, against a baseline taken before it was introduced.
Purchaser time per requisition and per quotation comparison.
Share of AI-matched lines accepted without change, and the correction rate.
Time to answer vessel status questions.
Number of price deviations caught before the order or the payment.
At the ship manager running Delta BI's Finance Automation, the workflow freed around 36% of purchasers' time. That result comes from the whole workflow: structured requisitions, one view of products and prices, and an AI layer that supports purchasers. We do not attribute it to AI alone, and we recommend you do not plan for AI savings without fixing the process first.
Next steps
Finance Automation combines the procure-to-pay workflow with an AI layer that supports purchasers. A walkthrough starts from one of your recent purchases, so you can see where support would save time and where people should stay in control.
Start with the decision
Tell us about your challenge. We will reply to arrange an assessment and identify a practical route to delivery.
Request a data assessment