AI for sales and business development in long, complex deals

Where does the time go in a complex deal?

In technology, government-facing and enterprise sales, a deal can run for months and involve a tender, several rounds of clarification and a committee on the client side. Most of the hours do not go into meetings. They go into the work around them: reading the client's strategy and past tenders, answering requirement after requirement, rewriting proposal sections that were written well last year, turning notes into actions, and updating the CRM so the forecast means something.

That is writing and analysis work, and it is where the evidence for AI is strongest. In a randomised study of 453 professionals, those using AI on professional writing tasks finished in 40% less time, with 18% higher quality (Noy and Zhang, Science, 2023). In a randomised study of 758 consultants, people using AI produced 40% higher-quality work and finished 25% faster on tasks AI handles well (Dell'Acqua et al., Harvard Business School working paper, 2023).

The same study carries the warning. On a task the AI handled badly, consultants using it were 19 percentage points less likely to be right. In sales, the tasks AI handles badly are the ones that matter most: reading a client's politics, judging what a decision-maker will accept, and deciding what the company can deliver. So the design question is which work goes to AI, and where a person checks it.

Which four workflows should a sales team build first?

In Cader Group's programs for sales and business development teams, these are the four we build first, on the team's own deals:

1. Account and tender research. AI reads the client's published strategy, annual reports, past tenders and news, and produces a one-page brief: priorities, likely evaluators' concerns, and open questions for the first meeting. The account lead checks it against what they know from the relationship.
2. Compliance matrices and proposal drafts. AI extracts every requirement from the tender document into a matrix, maps each one to approved past proposal content and flags gaps where nothing suitable exists. The bid owner fills the gaps and signs off every response.
3. Meeting notes to same-day follow-up. AI turns notes or a transcript into a summary, a list of actions with owners and a draft follow-up email. The account lead edits and sends it the same day, while the meeting is still fresh for the client.
4. Pipeline hygiene. AI compares CRM records against recent emails and notes, and flags stale stages, missing next steps and close dates that no longer match the conversation. The sales manager decides what changes.

Each workflow ends in a person's approval. On the Delegation Ladder, all four sit at Draft or Assist: the system prepares, and a named person decides.

What should never leave human hands?

Three things stay with people, whatever the tool can do:

• Pricing. Price reflects margin, competitive position and the relationship. AI can assemble the inputs; a person sets the number.
• Commitments. Delivery dates, scope, service levels and contract terms bind the company. Every commitment in a proposal needs a named owner who has confirmed it can be met.
• The relationship. Trust in a long deal is built by the people in the room. AI can prepare you for the meeting and help you follow up, but it should not speak for you to a client.

How do you keep client data confidential?

This is the risk most sales teams underestimate. In a 47-country study, 48% of employees said they had uploaded company information into public AI tools (University of Melbourne and KPMG, 2025; Indicative). In sales, that information includes client tenders, pricing and confidential requirements.

Three rules cover most of it:

• Approved tools only. Every workflow runs on tools the company has approved for client information, never on personal accounts.
• Read the tender's terms. Many tenders restrict how documents may be shared or processed. Check before uploading them anywhere.
• Approved content library. Proposal drafts draw on a curated library of past content cleared for reuse, not on every past bid, some of which carry another client's confidential details.

How do you measure the gain?

Take a baseline before the program on named tasks: hours to produce a compliance matrix, time from meeting to follow-up, share of CRM records with a current next step. Measure the same tasks afterwards. Track quality as well as speed, for example the number of requirements a reviewer had to correct. Do not measure win rates in the first months; a long deal cycle means the effect shows up much later, and many other factors affect it.

Measure where your own team stands: take the free Readiness Snapshot →