AI B2B Lead Generation Playbook

The playbook in one paragraph

Decide who you are for in rules precise enough for software to apply. Choose the events that make those companies likely to buy now. Build a list you can trust and classify it legally. Pick the channel that suits each segment instead of defaulting to cold email. Write from research, not templates. Get every response to a person the same day. Review weekly on pipeline, not activity. AI helps at every step, but it multiplies whatever is already there, so the order matters more than the tools.

Play 1: Write the ICP as rules

"Mid-sized companies that value innovation" cannot be applied by a person, let alone a model. Rules can. An illustrative example for a firm selling finance automation:

Illustrative ICP written as rules

Rule type	Rule	How it is checked
Include	UK-registered limited company, 50 to 500 employees	Company register and data provider
Include	Sector: wholesale, distribution or light manufacturing	Sector code, confirmed on the website
Include	Finance team of three or more	Job listings or data provider roles
Exclude	Already uses one of the named competing products	Technology signals, job ads
Exclude	Sole traders and ordinary partnerships	Legal form
Buyer	Finance director or head of finance; operations director as secondary	Title match, checked by a person

Then count the matches. The number tells you whether outbound is even viable, how many accounts each person can own, and how careful every message must be. A market of 1,500 companies cannot absorb careless volume.

Play 2: Choose signals before lists

A signal is an observable event that makes a company more likely to act soon: a new finance director, a funding round, a job advert for the role your product replaces, an expansion into a new site, a switch away from a competitor. AI is good at watching many sources for these and summarising them. Pick three to five signals that genuinely connect to your offer, and rank accounts by them. Contacting 200 companies with a reason beats contacting 2,000 without one.

Play 3: Build the list and check it

Enrich the matched companies with contacts, verify email addresses, and remove duplicates against your CRM before anything is sent. Then classify each record for UK rules. The ICO's guidance is that PECR's consent requirement for email marketing does not apply to corporate subscribers, but sole traders and some partnerships count as individual subscribers, and UK GDPR still applies to any named person. Every message must still identify the sender and offer an easy way to opt out. That classification belongs in the data, not in someone's head. This is a list of what to check, not legal advice; if your case is unclear, ask a qualified adviser.

Automating this step is where the time goes back. In AfM's published Allegiance Industries record, sales reps had been spending an estimated 15 to 25 hours a week on manual prospecting research before a daily automated system, built in 2026, took over company research, contact finding and CRM entry with duplicate checks. The client estimates it saves them $9,000 to $10,000 a year.

Play 4: Pick the channel for each segment

Matching channel to segment

Segment situation	Better first channel	Where AI helps
Buyers already search for the problem	Search ads and useful content	Keyword research, drafts, landing page variants
Small, high-value account list	Personal outbound and LinkedIn by a senior person	Account research briefs, draft notes
Large list, clear trigger events	Signal-led outbound email	Signal monitoring, drafting, reply sorting
Buyers who need convincing the problem exists	Paid social and a clear point of view	Creative variants, audience testing
Existing customers and past enquiries	Email to your own list and referrals	Segmentation, timing, draft follow-ups

Play 5: Write from research, not templates

Give the model a brief per account (the signal, the likely problem, one relevant proof point) and your approved positioning, and ask for a short first draft. A person then removes anything unverifiable and anything the recipient would find presumptuous. The target is a message the recipient can tell was written for their company, in under 100 words.

Play 6: Route every response the same day

Lead generation often fails after the lead arrives. In AfM's inbound processing record, follow-up had drifted to "1-3+ days" as volume rose, until an enrichment agent and priority notifications brought it to "Same Day". The lesson generalises: decide in advance who owns a response, how fast, and what information they receive with it.

Play 7: Review weekly on pipeline

Each week, look at the same few numbers per segment and channel: qualified leads, meetings held, opportunities created, and complaint and unsubscribe rates for any email. Make one change at a time and note it. Sends and impressions are inputs; they tell you work happened, not that it worked.

What to automate first

1. • List hygiene and enrichment (Play 3): high volume, easy to check, no reputational risk.
2. • Response routing (Play 6): cheap to build and it protects every lead you already pay for.
3. • Signal monitoring and research briefs (Plays 2 and 5).
4. • Drafting, with human approval.
5. • Sending volume, last, and only once the earlier steps hold.

For the compliance detail behind outbound, see automating lead generation without spamming. To price what the playbook produces, use the cost per lead calculator.

FAQ

How long should each play take to set up?

It depends on your data and team, so this playbook sets no timetable. A useful rule is to move to the next play only when the current one produces output someone has checked and would sign off.

Do I need all seven plays?

A company with strong inbound demand may barely need outbound, and Play 4 may send most effort to search and content. Plays 1, 6 and 7 apply to everyone.

Which play is most often skipped?

Play 1. Teams jump to tools and lists with a vague profile, then blame the tool when replies are poor. Writing the rules and counting the matches usually changes the plan.