The lead is qualified. Now make it easier to buy.

The campaign has worked. A credible prospect wants to talk. They have a problem worth solving, a reason to act and enough interest to give us their time.

Then we ask them to tell their story. Again.

First to the account manager. Then to pre-sales. An integration question needs checking. The proposal arrives with an assumption that needs correcting. The price follows a version behind.

Everyone is busy. The prospect is still waiting.

Somewhere along the way, buying from us has become another project to manage.

In my earlier article on agentic GTM, I explored how AI could connect customer signals with better qualification and commercial action. This is the next part of that story: turning a credible enquiry into a solution the prospect understands, the business can deliver and both sides can agree to.

AI could remove some of that effort for both sides. But the benefit depends on what happens between discovery, architecture, pricing and approval. Speed up one task while losing the context at the next handover, and the prospect still pays for the gap.

Don't make them start again

A useful handover should preserve what brought the prospect to us: their intended outcome, what they have already told us, the evidence behind qualification and the questions still open.

The discovery assistant can then begin with something helpful:

“You want engineers to spend less time finding maintenance information. You've mentioned three sites and a six-week pilot. Have I understood that correctly, and which outcome would make the pilot worthwhile?”

The prospect can correct the brief before we build on it. The next specialist receives that correction too. Their answer should travel further than the meeting in which they gave it.

Company research can make the conversation more relevant, provided we keep its status clear. An expansion announcement might prompt a question about additional sites. It cannot establish how many sites belong in the quote.

The same evidence should travel back into GTM. If discovery repeatedly uncovers an expectation the service cannot meet, product and marketing need to know. Otherwise, the next campaign will make the same promise and the next prospect will encounter the same disappointment.

Make every question earn its place

An AI assistant could make discovery feel more like a conversation with someone who has been listening. It can summarise an answer, identify something that does not add up and ask a follow-up because the response changes feasibility, scope or value.

Each question should earn the time it takes to answer.

“Which systems do you use?” may be necessary. “We need to establish whether current stock information is available, because that determines whether the assistant can help plan a repair” gives the question a purpose.

I would also let the prospect say “I don't know”, bring a colleague into the discussion and move to a person without starting again. They should be able to review the emerging brief, see outstanding questions and correct an assumption before it appears in a proposal.

Nobody should have to negotiate their way out of a chatbot conversation. The assistant needs a stopping point: enough information to recommend a next step, or a clear explanation of why a specialist needs to join.

The prospect should leave discovery understanding more about their decision than when they arrived.

The question that changes the solution

The fictional manufacturer in my Agentic Solution Studio wants an AI engineering assistant. It should find the right maintenance procedure, check parts availability and help prepare a maintenance request.

The prospect believes the reporting database already contains everything required. On that basis, connecting it to the assistant sounds straightforward.

Then we inspect the supplied evidence. The same equipment identifier appears at two sites. The manual's approval status sits outside the export. Stock information refreshes overnight. The shared reporting account tells us little about which information each engineer should be allowed to retrieve.

Now consider the question the finished assistant is expected to answer:

“Which seal kit do we need for Pump 104, and can we schedule the repair tomorrow?”

A convincing answer could still refer to the wrong pump, the wrong procedure or a part that is no longer available. Before we promise the capability, we need to resolve all three.

The pre-sales assistant has a useful question of its own:

“Does the first pilot need to confirm current parts availability, or would finding approved maintenance information be enough to test its value?”

The answer changes the offer:

The smaller pilot may be perfectly sensible. If the sponsor's real priority is planning tomorrow's repair, though, finding a manual faster only tests part of the idea. That trade-off belongs in the conversation before it becomes a disappointment at sign-off.

In the simulation, the connected option needs ten weeks against the prospect's six-week ambition. We can discuss a narrower scope, a later date or whether to proceed. Carrying six weeks into the proposal simply leaves delivery to explain the difference later.

Pre-sales is still scoping from supplied evidence. The paid engagement must validate the wider environment. An assistant should preserve that boundary rather than treating a sample extract as a completed assessment.

Help the prospect make the case in the room we're not in

The prospect usually has more people to convince. Engineering wants to understand dependencies. Finance wants to understand the commitment. Procurement wants clear terms. The sponsor needs to explain why this is worth doing.

Our contact has to carry the case into those rooms. The proposal should make that job easier.

Approved product and service collateral provides the starting material. The prospect-specific argument should connect their stated priority to a confirmed problem, the proposed capability and a measurable outcome. Unknowns need to remain visible.

For our manufacturer, finding information faster is a plausible pilot objective. Reduced production downtime would require further evidence. Removing that unsupported claim may make the proposal less dramatic. It also gives the sponsor a case they can defend.

A brief Statement of Work makes the commitment concrete: what will be delivered, what each side must provide, how success will be assessed, what is excluded and how changes will be handled. Delivery can assess the promise. The customer can see what they are agreeing to.

Pricing should follow a versioned catalogue and financial model. The agent can propose justified quantities and explain why an integration adds effort. Deterministic calculations should apply the rates, recurring charges, discounts and margin rules. A language model should not improvise a plausible-looking total.

The prospect sees the commercial worksheet. Internal delivery costs and margin support the seller's approval process. The executive slide summary should present the same scope and price as the detailed proposal.

Change the site count and the worksheet, proposal and slides should change together. The prospect should never have to ask which of our three prices is the right one.

Keep the promise intact through every handover

The joins between tasks are part of the product we need to engineer.

I would use one structured opportunity record containing confirmed requirements, source references, assumptions, unresolved questions, selected services, pricing inputs and approval status. Documents become views of that record.

AI is useful where interpretation matters: understanding an answer, identifying conflicting evidence, proposing the next question or explaining a recommendation. Workflow rules can handle calculations, required fields and version controls. Anthropic's distinction between predefined workflows and more adaptive agents is helpful here: choose flexibility where the task benefits from it, and justify the additional complexity through evaluation. Building effective agents

The catalogue needs owners too. Someone must maintain capabilities, exclusions, delivery assumptions and commercial rules. If the business cannot say clearly what it sells, how it delivers it and where the boundaries sit, the agent inherits that uncertainty.

Customer information must stay within the appropriate account and access boundaries. Prospect-supplied documents are evidence to examine, not authority to change pricing policy or approve commitments.

Human review should be concentrated where judgement or authority is needed. A technical exception needs the right specialist. An unusual discount needs the appropriate commercial approval. Reviewers should see the unresolved issue and its evidence without having to reconstruct the entire discovery conversation.

Once approved, a material scope or price change must clear that approval. The CRM should retain who approved which version and why. Creating a draft, approving it and sending it are separate events.

Done well, these controls settle questions while they are still inexpensive to resolve. They also give delivery a clear record of the promise it is expected to fulfil.

Follow the saving all the way to the P&L

The proposal is ready sooner. The team spends fewer hours assembling it. The business case is beginning to look attractive.

Now we need to follow those hours to a financial outcome.

Here, “cost of sale” means the commercial effort required to win business. That is a management measure whose boundaries we need to define with finance. It should not automatically be equated with the statutory cost-of-sales line. Pre-sales, sales operations, software and external support may sit in different expense categories.

We also need to separate time released, spending avoided and an actual reduction in expense.

Consider a separate, illustrative example of the seller's process. A team handles 100 comparable opportunities each quarter, averaging 12 hours of pre-sales, design and proposal work per opportunity. After introducing an assisted workflow, that falls to eight hours, including human review and correction.

That releases 400 hours. At an assumed fully loaded rate of £75 per hour, the capacity has an allocated value of £30,000.

That is £30,000 of capacity value. Finance still needs to know what changed in the spending.

If the same salaried team remains in place, payroll has not fallen. The capacity might support more opportunities, better technical evaluations or faster responses. Each could be valuable, but the resulting benefit still needs demonstrating.

Suppose the workflow adds £10,000 a quarter in incremental platform and external support expense. With no other change, the recurring expense base has increased by £10,000.

Now suppose the released capacity allows the business to remove £20,000 of existing quarterly contractor expense without reducing quality or shifting work elsewhere. The recurring expense reduction is then £10,000 after the new costs. The contractor reduction is one way the released capacity is realised; it must not be added to the £30,000 capacity valuation as another independent benefit.

There are still one-off implementation and integration costs to recover, with their P&L timing determined by the applicable accounting treatment. Avoiding a planned hire would instead be a saving against a forecast, rather than a reduction from today's spending.

A credible commercial case should be able to follow that trail: the work removed, the capacity released and the expense or profitable outcome that changes.

There is another possibility. Better scoping could reduce unpriced custom work and delivery rework after signature. That may protect delivery margin. Track it separately from acquisition efficiency, and check that apparent pre-sales savings have not simply moved effort into delivery.

Two hours of work can still mean three days of waiting

Proposal turnaround time tells us something useful. We need to understand what sits inside it.

I would track prospect effort, seller effort and the quality of the resulting commitment:

A proposal might need two hours of preparation and spend three days waiting for approval. A faster draft still lands in the same queue unless ownership and routing change. Measure working time and elapsed time separately so we improve the part that is actually holding things up.

Compare similar opportunities and include unsuccessful pursuits. For a cohort's pre-sales cost per win, include effort spent on its losses and no-decisions as well as its wins, allowing enough time for outcomes to emerge. That is only part of total customer acquisition cost; do not quietly leave marketing or account-management costs out of a broader claim.

Review rejected opportunities as well. A process that becomes quicker by excluding complex but valuable prospects may improve a dashboard while damaging the business.

Before expanding, compare the assisted approach with the existing process and a simpler structured workflow. Give it difficult cases: conflicting requirements, unsupported integrations, missing owners, stale collateral and a discount it has no authority to approve. Count the time people spend finding and repairing its mistakes.

Start with a decision you can improve

For a first implementation, I would choose one service family and a bounded set of discovery questions, catalogue rules and approval paths. Establish the baseline, then test whether prospects and staff actually have an easier time reaching a sound decision.

Look for the difference in the next conversation. Does the account manager understand the problem more clearly? Can the solution architect work from the brief without rebuilding it? Can the prospect explain the recommendation, its limits and the next decision to a colleague?

I've explored that process in Agentic Solution Studio, using a fictional manufacturer, synthetic evidence and a scripted workflow. It demonstrates the decision structure; it is not evidence of a live agent's performance or measured cost savings.

Try changing whether the prospect needs current parts information, then change whether data owners are available. Follow the effect through the recommendation, pricing, proposal and Statement of Work.

A faster proposal earns its keep when it helps the customer reach a well-informed decision and gives delivery a commitment it can fulfil.

We worked hard to earn the prospect's interest. The buying experience should give them a reason to keep it.