A sales team can now produce twice as many proposals. The technical director still checks every delivery promise, and the operations team still has the same capacity. If the additional proposals pursue the wrong customers, faster writing has created a larger queue and more ways to promise too much.
For a Norwegian technical-services business, I would judge AI in sales by the work it helps the company win and deliver profitably. The first useful application may be deciding which opportunity deserves attention, rather than drafting every request that arrives.
Start with the decision to bid
Imagine a hypothetical supplier of maintenance services receiving a tender for several sites. The customer needs a response time the supplier can meet near its existing team, but not at the most distant location. An assistant can extract that requirement from the documents. It cannot turn an unavailable technician into delivery capacity.
Before drafting, ask the commercial owner to approve a short opportunity note: customer need, mandatory requirements, delivery capacity, expected contribution and the most important uncertainty. AI can organize the evidence and point to the relevant pages. A person must decide whether to bid, seek clarification, partner with someone or decline.
Treat a missing answer as missing. A fluent summary that silently fills gaps is a poor basis for a commercial commitment. For public tenders, assess the specific tender rules and clarification process separately; this article is not a guide to public procurement law.
Separate source material from promises
Give the assistant a controlled set of current service descriptions, approved references and commercial terms. A previous proposal is evidence of what was offered in that situation, not proof that the same price, staffing or guarantee is available today.
In the maintenance example, assign three checks before release:
- The sales owner checks whether the proposal answers the customer's actual need and whether open questions are visible.
- The delivery owner verifies coverage, response times, staffing and technical claims.
- The authorized commercial approver checks price, exclusions, liability and other commitments against the company's mandate.
One person may hold several roles in a small AS. The decisions still need to be explicit. Do not let an assistant send a proposal or accept changed terms merely because it was allowed to prepare a draft.
Customer documents may contain confidential plans and personal information. Use approved tools and permitted material, and follow the guide to AI agents and GDPR in Norway. Remove unsupported references and claims, even when they would make the proposal more persuasive.
Run a sales experiment that can disappoint you
Choose one recurring proposal category and a defined trial period. Compare similar opportunities handled with and without AI support, or introduce the method gradually. Record why each opportunity was selected so a stronger customer mix is not mistaken for a technology effect.
Track total preparation and review time, material errors found, qualified opportunities pursued, wins and delivery contribution when it becomes known. Early indicators such as faster responses are useful, but a short trial may finish before contracts are awarded. Keep those outcomes open rather than counting pending offers as success.
Suppose, purely as an illustration, ten proposals previously led to three wins. AI makes twenty proposals possible, but only two turn into suitable contracts and senior review time rises. That is not automatically an improvement. Equally, fewer proposals with healthier margins may be a good result. Look at absolute contribution and the full cost of selling, not just a percentage win rate.
Keep the relationship where judgment matters
A useful automated draft gives the salesperson more time to understand a customer's uncertainty. It should not replace a necessary conversation with confident assumptions. In a small specialist market, a poorly grounded delivery promise can affect later discussions with the same buyer.
SSB identifies marketing and sales among common AI uses in 2026. That describes use, not demonstrated sales uplift. Set your own evidence threshold before expanding.
Leave the meeting with one changed decision
The CEO should select a proposal category, appoint the accountable owner and decide what would justify expansion or stopping. The board becomes relevant if the change materially alters commercial risk, investment or the business model, rather than the wording of individual offers.
When the customer's concern is security, use the guide to substantiating cybersecurity answers. When faster work changes the offer itself, use the pricing comparison. The practical objective is a proposal the customer can trust and the company can deliver.
Sources and scope
Sources checked on 12 October 2026. The maintenance supplier and proposal counts are hypothetical. The trial design and approval roles are my recommendations, not documented sales results or a prescribed procurement process.