{"name":"Bench notes","version":"1.0","sha256_of_posts":"88d19978389827b758a3070d56b1cad28048d6718904084b25fbc55b663071d9","full_posts":"posts.json","posts":[{"id":"internal-meeting-linked-to-account","title":"A meeting linked to an account may have nobody from that account in it","kind":"finding","status":"current","track":"data","needs":["call-data","context-read","local-config"],"applies_if":"Your engagement or relationship score counts meetings linked to an account by id.","skip_if":"Your score already requires an attendee from the account on every counted meeting."},{"id":"prior-contact-check-misses-sequencer-mail","title":"A prior-contact check on your CRM misses mail your sequencer sent","kind":"failure","status":"current","track":"data","needs":["crm-read","local-config"],"applies_if":"You use a sequencer or a separate sending domain and a prior-contact check reads only the CRM or the mailbox graph.","skip_if":"Every outbound tool logs each message to the CRM and your check already reads all of them."},{"id":"usage-limit-arrives-as-text","title":"A usage limit reaches a headless agent as plain text instead of an error","kind":"failure","status":"current","track":"agents","needs":["scheduler","agent-transcripts"],"applies_if":"You schedule headless claude -p jobs and judge a run by its exit code or by the last line of its log.","skip_if":"Every job already records a stop reason from the CLI output text, and a run with no real work counts as a failure."},{"id":"classify-agent-stoppages","title":"Most stops in an agent queue never needed the human","kind":"finding","status":"current","track":"agents","needs":["agent-transcripts","local-config"],"applies_if":"Your work queue holds agent handbacks to a person, and its waiting count grew week over week.","skip_if":"Your work queue holds fewer than 10 to 20 recorded agent stops, so no pattern shows yet."},{"id":"invented-because-passes-grounding","title":"An invented \"because\" passes a claim-by-claim grounding check","kind":"technique","status":"current","track":"data","needs":["agent-transcripts","context-read"],"applies_if":"Your job config runs a grounding check on summaries or alerts that explain why something happened.","skip_if":"Your agent transcripts show only lookups such as a field value or a list, with no explanations."},{"id":"re-logged-tasks-crowd-out-calls","title":"Re-logged call summaries crowd the calls out of an agent's evidence","kind":"finding","status":"current","track":"data","needs":["agent-transcripts","crm-read","call-data"],"applies_if":"Your jobs back-fill email or call activity into the CRM as tasks, and your agent's store also holds the call transcripts those tasks summarize.","skip_if":"Your ingest config reads call transcripts or CRM tasks, not both."},{"id":"closer-and-front-door","title":"More agents grow the backlog while a closing job and one capped channel shrink it","kind":"technique","status":"current","track":"agents","needs":["local-config"],"applies_if":"Your work queue is the board, ticket list or file where agents file items for a person to act on. It shows one approver for items that several jobs and agents file.","skip_if":"No work queue exists in your setup, or it holds items from one producer whose digest is your only entry point."},{"id":"assert-without-looking","title":"Agents assume before acting and claim after, both without checking","kind":"finding","status":"current","track":"agents","needs":["agent-transcripts"],"applies_if":"Your agent reports status such as done, live or verified to people in chat, tickets or digests.","skip_if":"Your agent's status reports come from a test that reads the system, never from the agent's own text."},{"id":"record-age-is-not-sync-lag","title":"A record's age is not the age of your sync","kind":"finding","status":"current","track":"data","needs":["agent-transcripts","crm-read","scheduler"],"applies_if":"Your agent or digest output tells sellers how current a CRM record is, such as a contact's title or an account's last activity.","skip_if":"Your agent transcripts show no comment on freshness, only field values."},{"id":"business-clock-in-the-answer","title":"Agents need the business clock in the answer instead of an optional tool","kind":"technique","status":"current","track":"data","needs":["agent-transcripts","crm-read"],"applies_if":"Your agents or digests answer this-quarter or last-week questions, and your CRM's fiscal year starts outside January or data is pinned to a past date.","skip_if":"Your CRM's fiscal calendar uses calendar quarters, its data is live, and all users share one timezone."},{"id":"default-to-run","title":"Most of my agents' questions shouldn't have been questions","kind":"finding","status":"current","track":"agents","needs":["local-config"],"applies_if":"Your work queue is the board, ticket list or file where agents file items for a person to act on. It holds agent-filed items waiting on your go-ahead, and the dispatcher's run log shows idle slots.","skip_if":"No work queue exists in your setup, or its waiting items are mostly finished work for review, not requests to start work."},{"id":"agent-judges-overbuild","title":"Have the agent judge overbuild instead of the human","kind":"technique","status":"current","track":"agents","needs":["scheduler","local-config","agent-transcripts"],"applies_if":"Your work queue is the board, ticket list or file where agents file items for a person to act on. It, or your transcripts, show agents proposing new jobs, hooks or scripts that a person approves.","skip_if":"No agent in your transcripts proposes new jobs, hooks or scripts. Or your approval log shows the approver rejecting agent proposals on scope in the last 90 days."},{"id":"invited-attended-mentioned","title":"Being invited or mentioned does not mean someone attended a meeting","kind":"finding","status":"current","track":"data","needs":["agent-transcripts","call-data","crm-read"],"applies_if":"Your jobs read calendar invites and call recordings for the same meetings, to back-fill the CRM or tell sellers who met whom.","skip_if":"Your only connected meeting source is a recorder that lists who actually spoke."},{"id":"no-volume-to-self-learn","title":"Most B2B teams lack the deal volume a self-learning signal model needs","kind":"finding","status":"current","track":"agents","needs":["crm-read"],"applies_if":"Your CRM or a connected vendor tool scores accounts or signals with a model that learns from your outcomes.","skip_if":"Your CRM shows thousands of closed-won and closed-lost opportunities a year."},{"id":"recommendation-would-change-if","title":"A recommendation has to show what would change it","kind":"technique","status":"current","track":"agents","needs":["agent-transcripts"],"applies_if":"Your decision log or transcripts show agent approval questions on most days, and the recommended option wins most of them.","skip_if":"Your decision log or transcripts show fewer than 5 agent approval questions a week."},{"id":"lessons-log-read-first","title":"A lessons log agents 'read at start' becomes write-only","kind":"finding","status":"current","track":"agents","needs":["agent-transcripts","local-config"],"applies_if":"Your agent instruction files tell agents to read a notes or lessons file before they work.","skip_if":"The lessons file fits in one read, and transcripts show sessions reading the file whole."},{"id":"gate-not-sharper-rule","title":"A rule that keeps getting broken needs a gate instead of a sharper rule","kind":"technique","status":"current","track":"agents","needs":["local-config","agent-transcripts"],"applies_if":"Your agent instruction files loaded at session start hold more than about 50 behavioral rules.","skip_if":"Your instruction files loaded at session start fit on one screen, about 50 lines in all."},{"id":"invisible-field-looks-empty","title":"CRM permissions make an invisible field look empty","kind":"finding","status":"current","track":"data","needs":["crm-read","local-config"],"applies_if":"A job reads or fills CRM contact fields through an integration user whose permissions an admin outside your team controls.","skip_if":"Your sync config reads only standard fields, through a system-administrator user that no one else manages."},{"id":"crm-amount-needs-its-currency","title":"A CRM amount is a number whose currency sits in a separate field","kind":"finding","status":"current","track":"data","needs":["crm-read","local-config"],"applies_if":"Your reports, digests or agents total CRM opportunity amounts, and multi-currency is turned on or you read more than one CRM.","skip_if":"Every opportunity in every CRM you read is in a single currency."},{"id":"rate-limits-not-permissions","title":"Keep an agent queue's speed caps separate from what a run may touch","kind":"technique","status":"current","track":"agents","needs":["local-config"],"applies_if":"A scheduler entry starts agent runs from a queue, and its config holds throughput caps for 2 or more agents.","skip_if":"No scheduler entry starts agent runs. Agents start only by hand, one task at a time."},{"id":"review-floor-fresh-eyes","title":"Every agent result gets one fresh-eyes review","kind":"technique","status":"current","track":"agents","needs":["agent-transcripts","local-config"],"applies_if":"Your work queue is the board, ticket list or file where agents file items for a person to act on. It lets agents mark their own results done.","skip_if":"No work queue exists in your setup, or it requires a person's approval before any agent result moves to done."},{"id":"memory-beats-docs","title":"Memory beats docs only when it's small and loaded by the work","kind":"technique","status":"current","track":"agents","needs":["local-config"],"applies_if":"Your agents load a memory or rules file into every session and it is nearing its load limit.","skip_if":"Your always-loaded memory or rules file is under half its documented load limit, or no such file exists."},{"id":"meeting-happened-is-evidence","title":"Whether a meeting happened depends on which source saw it","kind":"finding","status":"current","track":"data","needs":["crm-read","call-data"],"applies_if":"You back-fill meetings as CRM activity, or report meeting counts per rep, from more than one source: calendar, recorder or email.","skip_if":"Every meeting you count comes from one recorder that only creates a record when a call is recorded."},{"id":"go-live-same-session","title":"Flip an automation live in the session that proved the dry run","kind":"technique","status":"current","track":"agents","needs":["scheduler","local-config"],"applies_if":"Your scheduler holds a job that writes to a CRM, a chat tool or a published page.","skip_if":"Every loaded job writes only local files that nobody else reads."},{"id":"automation-inventory-drifts","title":"A hand-kept list of scheduled jobs drifts from what's actually running","kind":"technique","status":"current","track":"agents","needs":["scheduler","local-config"],"applies_if":"Your setup has a written job registry and loads jobs in 2 or more schedulers.","skip_if":"All your jobs live in one scheduler whose own console shows each job's live state."},{"id":"job-change-splits-one-person","title":"A job change leaves two records for one person or keeps their old job","kind":"finding","status":"current","track":"data","needs":["crm-read"],"applies_if":"A job marks contacts departed or dedupes people across sources by name, and it serves 1 record per person to reps or agents.","skip_if":"Your contacts carry a person-level id from an identity source that survives job changes."},{"id":"platform-under-you-drifts","title":"Tool renames silently break agent skills and allowlists","kind":"technique","status":"current","track":"agents","needs":["local-config"],"applies_if":"Your skill, prompt or allowlist files name tools from MCP servers or platforms that another party maintains.","skip_if":"Your MCP config pins every server to a fixed version that you upgrade by hand."},{"id":"tool-wording-sets-search-width","title":"A tool's wording and defaults decide how wide the agent searches","kind":"finding","status":"current","track":"agents","needs":["agent-transcripts","local-config"],"applies_if":"Your agent answers account questions for sellers through a search tool that can span accounts or mix raw sources with extracted facts.","skip_if":"Your search tool takes one fixed scope and one source type, with no optional modes."},{"id":"duplicate-accounts-incomplete-pipeline","title":"Duplicate accounts make pipeline answers confident and incomplete","kind":"finding","status":"current","track":"data","needs":["crm-read"],"applies_if":"You route cards or answer pipeline questions per account, and your CRM holds parent-child accounts or you read 2 or more CRMs.","skip_if":"No 2 account records in your CRM share a name, and you read only 1 CRM."},{"id":"sentinel-values-look-like-data","title":"Placeholder values in required fields look like real data","kind":"finding","status":"current","track":"data","needs":["crm-read"],"applies_if":"Your dedupe or enrichment matches people or accounts by name, or your digests average CRM amounts, and imports or integrations fill required fields.","skip_if":"Your CRM's import log and integration users show no writes to required fields."},{"id":"one-meeting-several-accounts","title":"One meeting can belong to several accounts","kind":"finding","status":"current","track":"data","needs":["crm-read","call-data"],"applies_if":"You back-fill meetings or emails into the CRM against one account per activity, and you sell through partners or to groups whose people share calls.","skip_if":"Every meeting record in your CRM has attendees from exactly 1 outside account."},{"id":"late-transcript-duplicates-the-meeting","title":"A late transcript turns one meeting into two","kind":"technique","status":"current","track":"data","needs":["crm-read","call-data"],"applies_if":"Your jobs back-fill meetings or call activity into the CRM from both a calendar and a call recorder.","skip_if":"Your CRM's meeting records come from one connected source only."},{"id":"date-only-activity-lands-a-day-early","title":"Date-only CRM activities land on the previous day","kind":"finding","status":"current","track":"data","needs":["crm-read"],"applies_if":"You back-fill email activity as CRM tasks, or build timelines and response-time metrics from tasks and events across time zones.","skip_if":"Every activity you read carries a full timestamp with a time zone."},{"id":"name-what-replaced-the-gate","title":"When you remove the human gate, name what replaced it","kind":"technique","status":"current","track":"agents","needs":["scheduler","local-config"],"applies_if":"A job config in your setup writes agent-researched records to a list or the CRM with no approval step by a person.","skip_if":"Every job config that writes agent-researched records has an approval step that a person completes."},{"id":"attribution-at-ingest-goes-stale","title":"Attribution captured at ingest goes stale when the CRM changes","kind":"technique","status":"current","track":"data","needs":["context-read","crm-read"],"applies_if":"Your store keeps account ids on calls or emails, and the CRM's field history shows reps re-parenting contacts, tasks or accounts.","skip_if":"Every read in your setup joins to the CRM live, and no derived account id is stored on calls or emails."},{"id":"eval-answer-keys-go-stale","title":"Eval answer keys over CRM data go stale","kind":"finding","status":"current","track":"data","needs":["local-config","crm-read"],"applies_if":"Your eval folder holds a saved set of expected CRM answers, such as account owners or employers, reused over weeks.","skip_if":"Your eval config reads a frozen snapshot of the data, not the live CRM."},{"id":"stated-reason-hypothesis","title":"A stated reason is a hypothesis rather than a rule","kind":"finding","status":"current","track":"agents","needs":["local-config","crm-read"],"applies_if":"Your scoring config takes reviewer verdicts as input, and fewer than a few hundred accounts carry a verdict.","skip_if":"Your CRM holds a few hundred or more closed-won and closed-lost opportunities to test rules against."},{"id":"open-cases-outlive-recency-windows","title":"Long-open cases fall outside recency windows","kind":"finding","status":"current","track":"data","needs":["crm-read","local-config"],"applies_if":"Your ingest or retrieval config reads only recent CRM records, cases or activity, and customer-health views or alerts rest on them.","skip_if":"Your ingest config loads full history for every object."},{"id":"stale-view-announces-its-age","title":"A dashboard should show when its data was last built","kind":"technique","status":"current","track":"agents","needs":["scheduler","local-config"],"applies_if":"A dashboard or page that sellers read is built from the output of 2 or more scheduled jobs.","skip_if":"A daily check reads each feeding job's run log and alerts on a missed run, and that check leaves a record each day."},{"id":"cost-fix-scope-decision","title":"A cost fix that cuts coverage changes what the job does","kind":"technique","status":"current","track":"agents","needs":["scheduler","local-config"],"applies_if":"Your git log shows a cost cut to a job that writes CRM records or feeds a list.","skip_if":"Your git log shows no cost cut to any job that writes records or feeds a list."},{"id":"prose-limits-dont-hold","title":"Asking a headless agent to read less didn't hold","kind":"finding","status":"current","track":"agents","needs":["agent-transcripts","local-config"],"applies_if":"A scheduled headless agent's prompt file tells it to stop reading at a limit, such as a watermark or a page count.","skip_if":"Every agent run in your setup is an interactive session, and the scheduler holds no headless agent job."},{"id":"design-for-decay","title":"Compute a fact's age when it's shown instead of in a refresh job","kind":"technique","status":"current","track":"agents","needs":["local-config"],"applies_if":"One person approves changes to a list the page shows, and the approval log shows gaps of several days between reviews.","skip_if":"The records' edit history shows several different people editing them on most days."},{"id":"archiving-makes-a-duplicate-look-fresher","title":"Archiving a duplicate makes it look fresher to last-writer-wins","kind":"finding","status":"current","track":"data","needs":["context-read","crm-read"],"applies_if":"Your dedupe or enrichment config merges contact or account records from several systems and breaks ties by the most recent modified time.","skip_if":"Your sources never archive or soft-delete records, or you keep one record per source with no merge."},{"id":"domain-is-not-company","title":"An email domain does not prove which company someone works for","kind":"finding","status":"current","track":"data","needs":["crm-read"],"applies_if":"A job matches people, emails or meetings to accounts by email domain, and the CRM holds more than a few hundred accounts.","skip_if":"Your CRM has no parent-child links, partner accounts or agency accounts, and each account uses one domain."},{"id":"delegation-fixed-cost","title":"A sub-agent has a fixed token cost, so small work is cheaper inline","kind":"finding","status":"current","track":"agents","needs":["agent-transcripts"],"applies_if":"Your transcripts show sessions that spawn several sub-agents for small repeated tasks, or reuse one agent across rounds.","skip_if":"In your transcripts, each sub-agent makes dozens of task calls on its own distinct task."},{"id":"human-profile-url-beats-vendor","title":"A rep-entered profile URL beats a vendor's perfect-score match","kind":"finding","status":"current","track":"data","needs":["crm-read"],"applies_if":"A job enriches titles or departures from a vendor, and a rep-entered CRM field holds profile URLs for some contacts.","skip_if":"Your contacts have no human-entered profile URLs, so the vendor match is the only identity you have."},{"id":"transcripts-dont-say-which-side","title":"Call transcripts name the speaker without giving their side","kind":"finding","status":"current","track":"data","needs":["context-read","call-data","crm-read"],"applies_if":"Your fact store or signal cards draw on call summaries from 2 or more recording or conferencing tools, attributed to companies.","skip_if":"Your call summaries quote speakers verbatim and never paraphrase into \"we\" or \"they\"."},{"id":"recurring-meeting-id-names-series","title":"A recurring meeting's join id names the series rather than a single meeting","kind":"finding","status":"current","track":"data","needs":["call-data"],"applies_if":"You back-fill call recordings or transcripts onto CRM or calendar meeting records by matching on join link or meeting id.","skip_if":"Your calendar holds no recurring meeting series with a fixed join link."},{"id":"shared-first-name-crosses-sides","title":"A shared first name sends the seller's words to the buyer","kind":"finding","status":"current","track":"data","needs":["call-data","crm-read","context-read"],"applies_if":"A job attaches call quotes to CRM contacts from speaker names, or matches people across sources by name.","skip_if":"Your transcripts carry a full identity from the meeting platform for every speaker."},{"id":"owner-of-record-not-a-book","title":"The account owner of record is not always a working book","kind":"finding","status":"current","track":"data","needs":["crm-read","local-config"],"applies_if":"You route alerts or tasks to sellers by the CRM account owner, or agents answer coverage questions from that field.","skip_if":"No account in your CRM is held by an inactive user or a default owner."},{"id":"finding-in-a-doc-changes-nothing","title":"A finding written in a doc doesn't change the system","kind":"finding","status":"current","track":"agents","needs":["local-config"],"applies_if":"Your setup has process docs with a notes or pain-points section, kept apart from the code or rubric the process runs.","skip_if":"Every process in your setup is code, and each process doc is generated from that code."},{"id":"cost-is-turns-times-context","title":"Agent cost is turns times context rather than payload size","kind":"finding","status":"current","track":"agents","needs":["agent-transcripts"],"applies_if":"Your transcripts show agent runs of 20 or more turns, or runs that spawn several sub-agents.","skip_if":"Your transcripts and code show only single model calls, with no agent loop that re-reads context."},{"id":"gate-or-rule-procedure","title":"How to decide whether an agent constraint is a prompt rule or a code check","kind":"technique","status":"current","track":"agents","needs":["local-config"],"applies_if":"A job in your setup has an agent generate more than one buyer-facing message or CRM value per run.","skip_if":"Your job config holds every output for a person's review, and the prompt file has under about 10 rules."},{"id":"left-the-company-many-writers","title":"\"Left the company\" has many writers and many meanings","kind":"finding","status":"current","track":"data","needs":["crm-read","local-config"],"applies_if":"A job marks contacts departed, or reads a CRM status field, checkbox or enrichment flag to end a person's tie to an account.","skip_if":"Departures in your CRM come only from one HR-fed field that no other process fills."},{"id":"agents-reread-their-own-output","title":"Agents re-read their own wrong output as ground truth","kind":"failure","status":"current","track":"agents","needs":["local-config","agent-transcripts"],"applies_if":"Your transcripts show agent runs reading a page, plan or table that an earlier agent run wrote.","skip_if":"Your job configs show every run rebuilding its output from raw sources, with no read of its earlier output."},{"id":"planner-never-types","title":"Use the top model to plan and review, and cheaper models to write","kind":"technique","status":"current","track":"agents","needs":["agent-transcripts"],"applies_if":"Your top-model session transcripts list a tool that spawns sub-agents, and some sessions reach their context or usage limit.","skip_if":"Your top-model session transcripts show fewer than 10 tool calls each."},{"id":"cap-swarm-fail-open","title":"Cap sub-agent spawns with a hook, and let the cap fail open","kind":"technique","status":"current","track":"agents","needs":["agent-transcripts","local-config"],"applies_if":"An agent's tool list or config in your setup includes a tool that spawns other agents.","skip_if":"No agent's tool list or config in your setup includes a tool that spawns agents."},{"id":"soft-delete-keeps-the-crm-stamp","title":"A soft delete may not move your copy's modified time","kind":"finding","status":"current","track":"data","needs":["crm-read","local-config"],"applies_if":"A job config in your setup reads the CRM incrementally by last-modified time, such as an incremental cleanup or an enrichment pass.","skip_if":"Your job configs reload every CRM object in full on each run, with no freshness gate."},{"id":"current-state-needs-a-date-filter","title":"'What's happening now' needs a date filter instead of a relevance store","kind":"finding","status":"current","track":"agents","needs":["agent-transcripts","context-read"],"applies_if":"Your agent's tool config points at a relevance-ranked search store, and transcripts show it answering current-state account questions.","skip_if":"Your agent transcripts show account questions only about whole history, never about what is happening now."},{"id":"two-forecast-category-fields","title":"Two forecast-category fields: one set from stage, one reps keep","kind":"finding","status":"current","track":"data","needs":["crm-read","local-config"],"applies_if":"Your digests, routing or agents read forecast category, and your CRM has a custom forecast-category field alongside the standard one.","skip_if":"Your CRM's opportunity object has no custom forecast-category field."},{"id":"same-name-no-shared-id","title":"Same name, same account, no shared id: one person or two?","kind":"finding","status":"current","track":"data","needs":["crm-read","local-config"],"applies_if":"Your dedupe or enrichment matches people across CRM, calendar and email sources by name and domain, where many records lack an email or id.","skip_if":"Every contact you hold carries a verified email or a stable external id."},{"id":"employment-is-a-time-window","title":"People who leave a company lose the link to their past work","kind":"finding","status":"current","track":"data","needs":["crm-read","context-read"],"applies_if":"Your CRM user object holds deactivated users, and their earlier calls and emails sit in your fact store.","skip_if":"Your reports and agents read only activity dated after your oldest active user joined."},{"id":"lint-agents-for-token-waste","title":"Review an agent job's token cost before its first run","kind":"technique","status":"current","track":"agents","needs":["scheduler","local-config"],"applies_if":"Your skills folder or scheduler holds a skill or job that calls a model more than once per run.","skip_if":"No scheduled job or skill in your setup calls a model, and your model use is sessions you watch."}]}
