AI for quotes and follow-up: fewer lost deals
How an incoming quote request becomes a quick-to-approve draft, and how to make sure no quote goes without follow-up.
AI-assisted quoting turns an incoming quote request into a draft you can check quickly, and makes sure no quote you send goes without a follow-up. A person still approves the price and the final wording. The machine takes over the typing, the searching and the reminders.
Where deals get lost after a quote request
For many SMBs, the bottleneck is not the number of leads but what happens after they arrive. A request comes in by email, form or phone, then waits days while someone digs up the price list, a similar quote and the free capacity. By then, the customer is often talking to a faster competitor.
Fast response mattered long before AI. A 2011 Harvard Business Review article already found that most companies were not responding nearly fast enough to online enquiries. Today part of that slowness can be automated.
The quiet loss: follow-up
The other quiet loss is follow-up: no owner, no reminder, or it feels awkward to write again. The quote does not end in a no, it just fades away.
From quote request to draft
A good system is a chain of clear steps, not one big AI. For each step, know the input, the output, and where a human decides.
- Intake and requirement extraction: the AI watches the quote inbox and website form; for phone requests, a colleague's note or the call transcript is the input. It builds a structured sheet: who wants what, how much, by when, and what is missing.
- Filling the gaps: if key data such as size, quantity or location is missing, the system drafts a follow-up question.
The quote draft
- Quote draft: the draft is built in your own template from the current price list, product descriptions and similar earlier quotes. Old quotes show which items belong together.
From approval to CRM status
- Human approval: the salesperson or manager reviews, edits and approves. No quote goes out without approval.
- Scheduled follow-up: after sending, the system suggests or sends a personalised reminder on a preset rhythm. It refers to the quote and the customer's questions, not a generic "just checking in" email.
- CRM status: every step is logged in the CRM: received, sent, followed up, won or lost. For a lost quote the reason is recorded too, because that is where the team learns the most.
Processing incoming documents, such as technical descriptions or drawings, is a separate field. We cover it on our document AI page.
The limit: pricing decisions stay with people
The AI can suggest a price from the list price and earlier quotes, but it cannot make the pricing decision. Discounts, special terms, payment deadlines and risky customers are business decisions that a person is accountable for. A good system therefore highlights the points where the draft differs from the usual, instead of hiding them.
The same goes for the tone of follow-up. A long-standing key partner gets a different message than a first-time enquirer, and a colleague who knows the relationship should judge that. Define per customer category where the system may send on its own and where it may only draft.
Why now
Agentic CRM features are no longer experiments. Microsoft made its Dynamics 365 Sales Qualification Agent generally available in October 2025, and by September 2026 its documentation lists several sales agents. One of them takes a deal through to closing, follow-ups included.
The wider picture points the same way. According to Eurostat, 20.0 percent of EU enterprises with at least 10 employees used artificial intelligence in 2025, up from 13.5 percent a year earlier. Of the companies using AI, 34.7 percent also applied it to marketing or sales.
Three metrics to track from day one
Before rollout, measure where you stand, or you cannot say later what the system delivered. Three numbers are enough:
- Quote turnaround time: the time from receiving the request to sending the quote, in hours or working days.
- Share of quotes with no follow-up: the percentage of sent quotes that got no further contact. The goal is to bring it close to zero.
- Win rate: the percentage of sent quotes that became orders. It moves more slowly, so compare at least a quarter of data.
Where the numbers come from
If statuses are kept consistently in the CRM, all three numbers appear without a separate spreadsheet. If you have no such data today, the first month's job is to capture that baseline.
How to start
The safest start is a narrow, measurable slice, for example only website-form requests for a single product group. Microsoft itself recommends testing its agent on a smaller scale first, and reviewing outgoing emails in a simulation. That discipline pays off with any tool.
Not everyone needs an enterprise CRM. If you would build the process around your existing email, price list and customer records, a custom solution may be the right path. We write about that on our custom AI systems page.
Four mistakes to rule out up front
- An outdated price list: if the price list is not in one place and up to date, the draft will be wrong too. The system does not fix that, it only multiplies the error faster.
- Full autonomy too early: in the first weeks, have a person approve every outgoing email, and loosen the rules only for categories that have proven themselves.
- Pushy follow-up: frequent, identical reminders damage the relationship. Fewer messages with real content work better.
- Careless data handling: check where customer data is stored and processed. Microsoft's own documentation notes that data used by its agents may be processed and stored outside the user's primary region.
The next step
If a quote takes days at your company, or you do not know how many sent quotes nobody followed up on, start with a short assessment. On our process automation page we describe how we map and automate processes like this, step by step.
Sources
- Oldroyd, McElheran, Elkington: The Short Life of Online Sales Leads, Harvard Business Review, 2011 (BYU ScholarsArchive)
- Microsoft Dynamics 365 blog: Dynamics 365 sets the bar for agentic sales qualification on new benchmark, December 11, 2025
- Microsoft Learn: AI agents in Dynamics 365 Sales, updated September 25, 2026
- Microsoft Learn: Set up and configure the Sales Qualification Agent, updated September 18, 2026
- Eurostat: 20% of EU enterprises use AI technologies, December 11, 2025
- Eurostat Statistics Explained: Use of artificial intelligence in enterprises
FAQ
Can the AI send the quote on its own?
Technically yes, but we do not recommend it. The approach that works is for the AI to prepare the draft and for a person to approve the price and terms before sending. For follow-up reminders you can loosen the rules gradually, per customer category.
What data does it need to produce a good quote draft?
An up-to-date price list in one place, product and service descriptions, your company's quote template, and ideally a batch of earlier quotes. The more orderly this base is, the fewer corrections the drafts need.
Do we need to adopt a new CRM?
Not necessarily. If you already have a CRM, the process can plug into it; if not, a simpler customer and quote register is enough to start. What matters is that the status of every quote is kept in one place.
How soon will we see results?
Quote turnaround time and the share of quotes with no follow-up can be measured within the first weeks. The win rate moves more slowly, so you need at least a quarter of data for a meaningful comparison.
Related service
Process automation
Brauniq automates business processes and connects systems: invoicing tools, CRM, online stores and spreadsheets.
Also related: Document processing AI
