Predictable Revenue in the age of AI agents: what still works and what to change

What the original playbook got right

Predictable Revenue described how a B2B team can create new pipeline on purpose instead of waiting for referrals. Its core ideas are simple: split prospecting, qualification, closing and account management into separate responsibilities; aim outreach at a clearly defined ideal customer profile; start conversations by asking for direction and relevance rather than pitching; and measure every step so the result can be repeated.

None of that has expired. Teams that struggle with pipeline today usually skip the same fundamentals: nobody owns prospecting, the target market is a vague list of industries, and success is measured in emails sent instead of qualified meetings held.

What changed since the book

The environment around the playbook moved. Inbox providers now filter bulk and low-relevance email much harder, so volume without targeting damages deliverability for the whole domain. Buyers research vendors on their own, on review sites, communities and LinkedIn, long before they reply. And the cost of research, personalisation and data cleaning has dropped sharply because language models can do a first pass in seconds.

The practical consequence: the bottleneck is no longer the number of messages a team can send. It is the quality of the list, the relevance of the first message and how quickly a reply turns into a conversation.

Where AI agents fit, role by role

• Prospecting research. An agent gathers company size, tech stack, hiring signals and recent news for each account, then flags who matches the ICP and why.
• List building and data hygiene. Agents de-duplicate contacts, check role relevance and enrich missing fields before anything reaches a sequence.
• First drafts. An agent drafts a short, specific opener based on the research. A person approves or edits it before it is sent.
• Reply triage. Every reply is classified as interested, referral, not now, not interested or unsubscribe, with a suggested next step and the CRM updated.
• Meeting preparation. Before a call, the agent summarises the account, the thread and likely objections for the person running the meeting.

In the language of the book, agents take over a large share of the SDR's preparation and administration, so the human time goes to judgement and conversations.

Where people stay in charge

Positioning, the offer and the decision about who is worth contacting are strategic choices, not tasks to automate. So is anything that represents your company in public. In our systems, agents propose and people approve: messages to new prospects, changes to targeting and any action that cannot be undone go through a human check.

This is also how quality stays high. When a reviewer rejects a draft, the reason becomes feedback for the next batch, which is the same feedback loop between SDRs and account executives that the original playbook recommends.

A two-week build, step by step

Week one is about the foundations: agree the ICP and the problems you solve for it, set up sending domains and authentication, build the first account list and write the core sequence across email, LinkedIn and phone. Week two connects the CRM, switches on research and reply-triage agents, runs a small first batch and checks every reply by hand.

From then on the system runs, and the work becomes improvement: which segments reply, which messages convert replies into meetings, which channels are worth more effort. That weekly loop is where predictability actually comes from.