Use cases
Eight GTM use cases for the Amdahl learning model, from outreach to win/loss, each mapped to the calls that serve it today, what your code gets back, and where to start. Outreach needs no connected data.
The Amdahl learning model serves eight GTM use cases: outreach, positioning, ads, content, sales call prep, account-based expansion, churn risk, and deal research and win/loss. Outreach is one of them, and the only one that needs no connected data. Each row below names a job, the calls that serve it today, and what your code gets back.
Outreach
Amdahl writes and optimizes email and LinkedIn messages. Your agent sends what it knows about the recipient, and Amdahl returns the message, evaluated before anything ships.
| Job | Calls | You get back | Needs your data | Start with |
|---|---|---|---|---|
| Write one message from what you know about the recipient | The optimize call (messages.optimize) | One message | No | Optimize a cold email |
| A whole campaign: one message per recipient, from a CSV of what you know | The optimize call, once per recipient, or a batch | One message each | No | Optimize a campaign |
| The writer behind an AI SDR or any code that sends | The optimize call between the code that knows the recipient and the send queue, with a fallback | One message, or a fallback | No | Optimize before you send |
| Keep the sender's voice and house rules in every message | The optimize call with context (facts, rules, voice_examples) | One message | No | Keep your voice, rules and facts |
| Evaluate every claim against your customers before a message goes out | The optimize call with evidence: "workspace", which lists them in unsupported_claims | One message and the claims nothing backs | Yes | Evaluate claims against your customers |
| Grounded outbound: a first touch in your buyers' own words | Search (semantic, hydrate: true) for the pain, then the optimize call with evidence: "workspace", then an evaluation of the result | Quotes, then a message, then an evaluation | Yes | Onboarding skill, use case A |
| Evaluate a draft: every criterion passed or failed, with a reason for each | The optimize call with mode: "grade" and evidence: "workspace" | An evaluation | Yes | Evaluate a cold email |
| Optimize before it ships: evaluate what an AI SDR wrote before it sends | The optimize call with mode: "grade", read on grade.passes_bar | A send or hold decision | No | Optimize before it ships |
| Sequence review: every step of a live sequence | One evaluation per step, then read the criteria across steps | One evaluation per step | No | Evaluate a cold email |
| Prompt tuning: improve the prompt your team or agent writes from | optimize_prompt in improve mode, then get_prompt_run for the held-out lift | A revised prompt, only when it wrote better email | No | Improve a prompt |
| Writing-quality trend across a team | Many evaluations, logged and aggregated by you | Your own trend line | No | Evaluations as a trend |
Positioning
| Job | Calls | You get back | Start with |
|---|---|---|---|
| What closed-won accounts pushed back on | Search (filter on deals), then Search (semantic, scoped to those companies) | Quotes from the accounts that bought | Cohort, then voice |
| How customers describe the problem, in their words | Search, mode: "semantic" with hydrate: true | Quotes with their source | Search |
| Whether anyone named a competitor or a feature | Search with the term in double quotes | Every exact mention | Search modes |
Ads and content
Amdahl optimizes email and LinkedIn messages today. For ads and content, it gives the agent that writes them your buyers' own words.
| Job | Calls | You get back | Start with |
|---|---|---|---|
| Buyer language for an ad or a post | Search, mode: "semantic" with hydrate: true, on the pain or the outcome | Quotes with their source, for your agent to write from | Search |
| The proof points customers actually repeat | Search, filter on closed-won deals, then semantic on those accounts | Quotes from customers who bought | Cohort, then voice |
Sales call prep
| Job | Calls | You get back | Start with |
|---|---|---|---|
| Prep for a meeting with an account | One research job on the account | One cited answer | Research |
Account-based expansion
| Job | Calls | You get back | Start with |
|---|---|---|---|
| Accounts in a segment or stage, and what they said | Search (filter on accounts or deals), then Search (semantic, scoped to those companies) | A list of accounts, then their quotes | Cohort, then voice |
| Everything known about one account before you reach out | One research job on the account | One cited answer | Research |
Churn risk
| Job | Calls | You get back | Start with |
|---|---|---|---|
| What current customers said about pricing, a competitor or a gap | Search (semantic, scoped to customer accounts), and exact terms in double quotes for names | Quotes with their source | Search |
Deal research and win/loss
| Job | Calls | You get back | Start with |
|---|---|---|---|
| ICP and win/loss research that needs synthesis over your own conversations | One research job on the question | One cited answer | Research |
| Funnel and pipeline questions: counts, win rates, cycle time by segment | Search, mode: "filter" with group_by and metrics | Numbers, with the SQL that produced them | Search |
What every use case but outreach needs
Two limits apply to every row that reads your processed customer data. That is every row on this page except outreach evaluated without your data.
- It needs synced conversations. A workspace with none returns empty searches, an evaluation's four evidence criteria abstain, and an optimize call with
evidence: "workspace"falls back to an evaluation without your data. CheckGET /search/overviewfirst. - CRM presence is not a commercial relationship. CRMs create company records from inbound mail, so your data holds vendors, recruiters and press alongside buyers. For ICP, voice-of-customer and win/loss questions, scope to accounts with at least one call or deal, and say that you did.