Sales prospecting
B2B prospecting with generative AI: an assistant, not an autopilot
What generative AI actually changes about outbound prospecting at an SME, what it must never do alone, and the rules to follow before the first send.
AI audit & implementation · Small and mid-sized companies
More clients.
Better-informed decisions.
Less manual work.
I help small and mid-sized companies find where generative AI really changes the business, including where nobody expects it: analysing sales and stock, keeping an eye on competitors, building the missing tool in a few weeks, as much as winning and following up clients. The expected gain is agreed before we start, then measured in real conditions.
- 5-day sprint · €4,500 excl. VAT
- Diagnostic eligible for Bpifrance funding
- Fractional head of data & AI
5 minutes, free, no sign-up
A ten-day diagnostic on one area, with a costed plan at the end, client or not.
First system in production in four to six weeks.
What a project changes
- Requests and data coming in from everywhere.
- And nobody has time to make use of them.
- We start by looking at what really happens.
- Then we build the tools that turn them into leverage.
- More clients
- Better-informed decisions
- Less manual work
Client results
What changes, what it earns, and how it is measured.
Four transformations delivered for real clients. For each one: what changes in the daily work, what it lets the company do, and the measures you can judge it by.
- 01 Revenue
The website brings in enquiries every day.
20-person consultancy measured since March 2026See the context and the workWhat it lets the company do
The company has selling opportunities it did not have before, instead of depending entirely on its network and on referrals.
What I report as the measure
- Inbound enquiries: from one every three days to two a day
- One inbound enquiry in four turns into signed work
- Daily traffic up fifteenfold, from 10 to 150 visitors
- 02 Revenue
Proposals go out on the day of the meeting.
20-person consultancy running since 2026See the context and the workWhat it lets the company do
Sales follow-up becomes responsive: the proposal lands while the prospect still remembers the conversation, and usually before a competitor sends theirs.
What I report as the measure
- Time to send: same day, instead of the end of the week
- Preparation time: two minutes of review instead of an hour of writing
- Win rate on the proposals sent
- 03 Hours freed
Accounting and the CRM stay in step.
20-person consultancy checked monthly since 2026See the context and the workWhat it lets the company do
Teams stop rekeying the same data and work from one consistent record, instead of reconciling two tools that disagree.
What I report as the measure
- Data-entry hours removed each month
- Discrepancies between the two tools, flagged automatically
- Lag between an entry and its counterpart
- 04 Cash recovered
Every invoice is matched against the contract governing it.
Payback Group engagement run from 2022 to 2024 Earlier engagementSee the context and the workWhat it lets the company do
Discrepancies become visible before payment, across the whole flow rather than a sample drawn at random.
What I report as the measure
- Amounts flagged, then confirmed by the purchasing team
- Amounts actually recovered or avoided before payment
- Share of the flow checked: 100%, instead of a sample
Three of these four results come from the same client, a 20-person consultancy, in different areas of its business. The fourth is an engagement that predates the current offer.
All twelve buildsBeyond sales
Leverage in places you wouldn't expect.
Generative AI is not just for writing emails faster. For my clients, some of the clearest gains came from places nobody was looking: stock levels, supplier invoices, the data sitting unused in the ERP.
- Delivered for a client
Internal data analysis
Why has this product line been losing margin for three months?
An assistant connected to your sales, accounts and CRM answers in minutes the question that used to take a week of spreadsheets, and cites the figures it relies on.
An assistant that cites its sources - Delivered for a client
Sales, stock and logistics
How much to order, of what, and when?
Reorder points and quantities are worked out item by item from past sales, instead of thresholds set by hand. 36,000 items are managed this way at a parts distributor.
36,000 items managed - To build for you
Competitive intelligence
What are my competitors preparing?
An agent follows their websites, prices, job ads and announcements every week, then sends you a two-page summary: what changed, and what it may mean.
- Delivered for a client
Custom web tools
And that spreadsheet nobody dares to touch?
The application your business is missing, built with your team in a few weeks, production-grade, monitored, and hosted wherever you want. It belongs to you.
A business tool in weeks - Delivered for a client
Purchasing and invoicing
Are we really paying what was negotiated?
Every supplier invoice is checked against the prices, discounts and terms in the contract, across the whole flow rather than a sample. Errors are stopped before payment.
100% of the flow checked - Delivered for a client
Customer voice
Why do our customers keep calling back?
Analysing every conversation, not a sample, shows where repeat calls really come from. In one customer service centre, the cause was the system, not the advisers.
45,000 conversations analysed
These are only examples. The diagnostic exists to find the ones that matter in your company, including ones that are not listed here.
Find mineDo you recognise this?
Five places in the business where a tool changes the game.
Each is treated as a subject in its own right, with its typical builds, its timeframe and its own data privacy question. If one of them describes your week, that is probably where to start.
Demonstration
From the client meeting to an up-to-date CRM, with no rekeying.
The sales follow-up build, laid out as it actually runs. On the right, an anonymised extract of what the system produces after a meeting.
- 01
Client meeting
The sales meeting is recorded, with everyone’s agreement.
- 02
Notes and proposal drafted
The transcript becomes meeting notes, a follow-up email and a costed proposal, priced from the company’s real catalogue.
- 03
Human sign-off
Stop pointWhoever ran the meeting reads it, corrects what needs correcting, then sends it. Nothing goes out on its own.
- 04
CRM up to date
Contacts, companies and deals created and updated as a side effect, with no end-of-week data entry.
- Read the sales follow-up case
Commercial proposal Anonymised extract
Location study — catchment area
- Addressed to
- name withheld
- Meeting
- Tuesday 10 March, 2.30 pm
- Drafted
- Same day, from the transcript
Line items priced from the catalogue
- Catchment area study
- Competitive analysis over the agreed perimeter
- Forecast revenue estimate
- Write-up and presentation to the partners
Total excl. VAT
Reviewed by the salesperson, sent on the day of the meeting.
Where to start
Four doors, and they do not cost the same.
The first is free and you do it alone. The next two are short fixed-fee engagements: a prototype to judge on evidence, or a diagnostic to decide the plan. The last is a conversation, if you already know what you need.
5 minutes Free
Self-assessment
Find your priorities: ten questions, and your top three levers ranked.
The result appears without signing up. If none of it speaks to you, you have lost five minutes.
Find my priorities 5 days €4,500 excl. VAT
Prototype sprint
A prototype running on your own data within a week, so you decide on evidence, not on promises.
A data extract, a one-hour scoping call, a demonstration on Friday. You keep the prototype and the costing of what comes next.
What the sprint covers 8 days €10,000 excl. VAT
Operational diagnostic
Decide what to build, on what budget and for what expected gain — costed, on one area of your business.
Three half-days with your teams and read access to your tools. You leave with the plan, client or not.
Eligible for 40% funding by Bpifrance (Diag Data IA) for French SMEs of 10 to 2,000 staff: the company files the request, subject to approval.
What the diagnostic covers 30 minutes Free
Talk about a project
You already know what you need: let us talk about it directly.
No diagnostic imposed first. If it is not a job for me, I will say so straight away.
Book 30 minutesExtract from the deliverable
What the diagnostic actually buys you.
The template for one line of the plan, filled in on an example. There is one per lever identified, ranked by gain and by effort.
The figures above illustrate the template. Yours come from the measurement done at your company, not from a rate card.
Diagnostic · 8 days Template filled in on an example
- Process observed
- Preparing and sending proposals after a sales meeting.
- Cost today
- About 1 hour 10 minutes per proposal, 18 proposals a month, so roughly 21 hours monthly.
- Potential gain
- About 17 hours a month, and the proposal going out on the day of the meeting instead of at the end of the week.
- Cost to build
- Four to five weeks, for a system in production wired into the existing CRM.
- Assumptions
- Proposal volume stays stable, the price catalogue is current, human review is kept before anything is sent.
- Recommendation
- Start here: it is the best ratio of expected gain to effort required.
Data privacy
Your data does not leave without you knowing where it goes.
It is the first objection directors raise, and it is a fair one. It has precise technical answers, and which one applies depends on how sensitive your material is.
How I handle the question 01
One sensitivity scale, four different answers
From a programming interface under contract through to an open model running on a machine you control, depending on what you handle.
02
The questions asked before anything is built
Where the data is processed, whether it is kept, whether it trains a model, which subprocessors see it, and for how long it is retained.
03
What stays with you, spelled out
The architecture separates what may leave from what must not. Plenty of builds run on cut-down or anonymised data.
10+ years in the field
Data science & AI in production
Who you are dealing with
Anthony Crowther-Alwyn.
Ten years of data science inside large French companies, then AI systems put into production at software publishers, retailers and customer contact centres.
Since March 2026 I have run generative AI adoption projects at my main client, a 20-person consultancy, and at a remote monitoring company. The results shown on this site come from that work.
"What I learned there fits in one sentence: the obstacle is almost never the technology, it is knowing what to build first."
Guides
What I know, written down.
Admin and finance
Connecting your accounting and CRM without double entry
What to synchronise between accounting software and CRM, in what order, where generative AI adds something, and why monitoring decides everything.
Customer support
Customer service AI: listen to every call before you automate
Why a chatbot bolted onto a misunderstood customer service function makes things worse, how to analyse every exchange, and what to automate next.
Next step
Where would you start?
Start with the self-assessment: five minutes, free, and you leave with your top three levers. The ten-day engagement described alongside comes next, if you decide to go for it. If AI is not the right answer, I will tell you so.
5 minutes, free, no sign-up
Diagnostic
Paid engagement8 days
€10,000 excl. VAT · fixed fee, one area of the business
- A map of where the time actually goes in the area audited
- Each lever costed in hours freed, ranked by effort and by gain
- The data note: what leaves your systems, and where it is processed
- A three-month plan, naming the first system to build