AI Prompts for Outreach Angles and Message Hypotheses
The angle is the reason this message exists for this person now. It comes from research — a trigger, a stated priority, a visible pain, a peer's result — and it is what separates relevant outreach from a template with the company name changed. Most sequences fail at the angle, not the copy: the writing is fine and the reason to read it is absent.
These prompts derive angles from what you know about an account, phrase them as message hypotheses that can be tested, and run the test across a segment. They sit between account research and sequence writing. The free workflow 'Build better outreach for a real prospect' runs the same logic end to end for a single prospect.
Before you use these
Have these ready to replace the highlighted [variables]:
- Account brief or the facts you know about the prospect
- Your value drivers and two or three proof points with numbers
- The persona's likely priorities and what they are measured on
- Past messages that worked and did not, if you have them
The prompts
- 1. Derive angles from the research
- 2. Write message hypotheses to test
- 3. Run the angle test across a segment
1. Derive angles from the research
Act as an outbound strategist deriving outreach angles for [prospect name, role] at [company]. What we know: [facts from research with sources: triggers, priorities, hires, tech, statements] Our value: [problems solved, outcomes, proof points with numbers] Persona: [what this role is measured on, what they care about, what they are tired of hearing] 1. Generate 5–7 candidate angles. Each: the specific fact it rests on → the problem or opportunity it implies for this person → the bridge to what we do → the proof that makes it credible. 2. Specificity test: for each angle, could it be sent to a different company unchanged? If yes, discard or rewrite. 3. Rank by relevance to the persona's priorities and evidence strength. 4. For the top three: the one sentence that opens the message, and the one question that ends it. 5. Angles to avoid: the ones that rely on assumptions the research does not support, or that reference sensitive events (layoffs, litigation). 6. Sequencing: which angle leads, which follows if the first gets no reply, and why the order. Do not use flattery, 'I noticed you…' openers about trivial details, or feature lists as angles.
2. Write message hypotheses to test
You are converting outreach angles into testable message hypotheses for [segment / persona]. Angles: [ranked list with facts and bridges] Baseline: [current reply and meeting rates for this segment] Sample: [accounts available per hypothesis] For each angle produce a hypothesis card: 1. Hypothesis: 'Leading with [angle] to [persona] at [segment] will produce a reply rate of [x]% (baseline [y]%) because [reason].' 2. Message skeleton: opener referencing the fact, the problem statement, the proof, the ask — under [word limit], with the variable fields marked. 3. Success threshold and the minimum sample to read it (state the arithmetic: at baseline [y]% and a target lift of [z], roughly [n] sends are needed to see a difference that is not noise). 4. What is held constant (channel, timing, sender) so the test isolates the angle. 5. Failure learning: if the hypothesis fails, what it tells us (angle irrelevant, persona wrong, proof weak) and the follow-up test. 6. Risk: any way the message could damage the account relationship if it misfires. Keep hypotheses independent. If two angles would be tested on the same accounts, sequence them rather than running both.
3. Run the angle test across a segment
Act as an outbound program analyst reading the results of an angle test. Results: [per hypothesis: sends, delivered, replies (positive/neutral/negative), meetings, sample of reply text] Baseline: [segment baseline rates] Design: [what was held constant, duration] 1. Rate table per hypothesis: reply rate, positive reply rate, meeting rate — with the sample size and a plain statement of whether the difference from baseline is large enough to trust at that sample. 2. Reply content: the objections, questions and language prospects used — grouped, with what they reveal about the angle's relevance. 3. Confounds: anything that differed between arms besides the angle (timing, sender, list quality). 4. Decision per hypothesis: scale, iterate (with the specific change), or kill — with the reason. 5. The next test: the highest-value follow-up hypothesis and its design. 6. What to feed back to research and ICP: any segment or persona finding that changes targeting. Do not declare a winner from a handful of replies. State the sample needed if the result is inconclusive.
Related prompts
- Account Research
- Cold Outreach Sequences
- LinkedIn Outreach
- Sales Follow-Up Emails
- Buying Signals and Trigger Event Analysis
- Prospecting Strategy
- Free workflow: Build better outreach for a real prospect
Logical next step
After this, most sales teams move on to Account Research.
All Sales prompts · Search the full library
Want the free B2B Sales AI Starter Kit? Nine prompts as a target → engage → qualify workflow, delivered by email. See what's inside
✓ On its way — check your inbox in the next few minutes.
Send me this starter kit and occasional useful AI workflow updates. Unsubscribe anytime. Privacy Policy.