One rep. One week of prospecting. Run two ways.
Find 50 likely adopters, remove owned accounts, research 15, draft 10 emails, load 5 into Salesforce. Below: what it costs a rep by hand, and what it costs when Claude or ChatGPT does the clerical work across Oystercatcher, Salesforce, Apollo and email.
Tool counts, hand-offs and citations are structural: walk the seven steps and you will count the same. Time ranges are an illustrative estimate for a careful rep, not a measured customer result.
Your rep types this on Monday morning.
Name the device, not “a device”: the more specific the ask, the better the fit scoring and the signals. Six instructions. The assistant plans the steps, asks each connected tool for what only that tool knows, and comes back with finished work and the source of every fact. Nothing is sent or written to Salesforce until the rep says so.
Prompt · Claude or ChatGPT“Find the 50 interventional cardiologists most likely to adopt our transcatheter mitral valve repair system. Exclude accounts already owned by reps. Research the top 15. Identify the three strongest buying signals for each. Draft outreach for the top ten. Put the five highest-priority accounts into Salesforce and assign them based on territory.”
Seven steps. Same job, two ledgers.
- By hand01Find 50 likely adopters
Filter a list vendor by specialty and state, open CMS utilization files in a spreadsheet for volume, cross-reference by hand. Most reps skip the volume step.
3 tools open~3–6 hAgent-firstOystercatcherOne Oystercatcher search: cardiology, high Part B volume in the relevant procedure families, early-adoption signals, your geography. Every match scored 0–100 with the factors behind it.
A ranked table of 50. Ask “why is #7 above #12?” and the answer comes from the score.
a few minutes - By hand02Remove owned accounts
Paste 50 names into Salesforce one at a time, or export a report and VLOOKUP. Miss the ones filed under a different practice name.
2 tools open~45–90 minAgent-firstSalesforceOystercatcherLooks up each of the 50 in Salesforce, drops any with an owner, and backfills from the next-ranked matches so you still end with 50.
The same table with an “owned by J. Ortiz” note on each removal.
seconds - By hand03Research the top 15
Google each one: practice site, LinkedIn, PubMed, conference programs, local news. Notes in a doc, sources not kept.
5 tools open~5–10 hAgent-firstOystercatcherRuns Oystercatcher’s research on all 15 in batches. Each dossier covers procedure mix, publications, conferences, practice news, hiring and technology mentions, with a source link on every claim.
Fifteen dossiers, same depth for the fifteenth as the first.
a coffee break - By hand04Three buying signals each
Re-read the notes and pick. Quality depends on how tired the rep was on physician #11.
1 tool open~1–2 hAgent-firstRep approvesReads the 15 dossiers against your device’s adoption profile and picks the three signals that matter most for each physician.
A 15-row table: physician, three signals, a source for each.
under a minute - By hand05Contacts for the top ten
Search the enrichment tool per person, copy emails into a sheet, guess the practice administrator’s name.
2 tools open~1–2 hAgent-firstOystercatcherApollo / ClayVerified physician email from Oystercatcher where it exists; the people around the account — administrator, lab manager, health-system parent — from Apollo or Clay.
A contact block per account with confidence shown.
a minute - By hand06Draft ten emails
Write each from the template, personalize from the notes doc. Ten tabs open.
2 tools open~1.5–3 hAgent-firstEmailRep approvesTen short emails, each opening on one sourced signal, saved to the rep’s Drafts folder. Nothing sends.
Ten drafts in the rep’s voice, waiting to be edited and sent.
a minute - By hand07Load five into Salesforce by territory
Create each record by hand, look up the territory map, set the owner, then remember to update the targeting tool too.
3 tools open~30–60 minAgent-firstRep approvesSalesforceOystercatcherShows the five records and the owner each gets under your territory rules. On approval, creates them in Salesforce and assigns the same leads in Oystercatcher so both systems agree.
Five accounts with the right owners, and a summary linking back to every source.
a minute, plus approval
Every signal comes with its receipt.
After step four the assistant returns one row per physician: a fit score from Oystercatcher, three buying signals, and the document each one came from. A manager can ask “why these five?” and get an answer that survives a pipeline review.
Medicare Part B utilization by NPI. Not in a CRM, not in Apollo, not something a chatbot knows.
Practice ownership and affiliation: owner-operator, group partner, or employed under a value-analysis committee.
A 0–100 score with named factors and a dossier where every finding carries its source.
Research is complete on all 15. Three strongest adoption signals for the top three; the full table follows.
- Top-decile Part B volume for the target procedure family, three years running. [CMS Part B]
- Moderated a 2026 panel on next-generation structural heart devices. [conference program]
- Practice announced a second cath lab opening this fall. [press release]
- Physician-owned: capital decisions sit with the partners, not a health-system committee. [NPI + practice data]
- Co-authored a case series on a competing platform, limitations noted. [PubMed]
- Two interventional hires posted last quarter. [careers page]
- Early adopter of the prior device generation, per 2024 procedure-mix shift. [CMS Part B]
- Site investigator on an active device trial. [ClinicalTrials.gov]
- Quoted on expanding the structural heart program. [local news]
Draft outreach for the top ten now, or adjust the ranking first?
Illustration · sample dataThe assistant is smart. It still has to know who the doctors are.
Run the same prompt with Claude connected only to Salesforce, email and Apollo, and step one fails quietly. Nothing in those tools says which cardiologists perform your procedure, how often, in what kind of practice, or who adopted the last platform early. You get a plausible list of cardiologists from the web. Not the right 50.
Apollo and Clay know companies and job titles, which is what you want for the practice administrator’s email. Clinical fit lives elsewhere: Medicare procedure volumes by NPI, practice ownership, state-board records, publications, conference activity. That is what Oystercatcher supplies, as a ranked, explained, cited answer. Take it out and the workflow still runs. It runs on guesswork from the first step.
Less clerical work in front of the call. Better first calls.
The rubric does not get tired. The fifteenth dossier is as thorough as the first, across every rep on the team.
Each one is tied to a document. “Why these five?” gets the score, the factors and the source, not a shrug.
Salesforce and the targeting tool are written in the same run after one approval. Neither goes stale.
Emails land as drafts. Salesforce writes are previewed and wait. The rep stays the rep.
Salesforce creates and email sends are previewed and wait for a person. That is how the assistants are designed, not a limitation.
Searches, research runs and contact lookups draw on your Oystercatcher plan. The assistant reports remaining quota before spending it.
Claude and ChatGPT connect to Salesforce, Gmail, Outlook, Apollo and Clay through those products’ own connectors. Oystercatcher’s native Salesforce sync is coming soon; this workflow does not depend on it.
Oystercatcher is built on public professional data — the NPI Registry, CMS datasets and state boards. It returns no protected health information, and no Open Payments fields through the connector.
Manual time ranges are estimates for a careful rep. Measured figures will be published when a referenceable customer has recorded them.
Two minutes. No IT ticket.
One server URL works in every assistant that speaks MCP — the open Model Context Protocol behind Claude’s and ChatGPT’s connectors. Each rep connects with their own Oystercatcher sign-in, so access is per user, revocable from Settings, and every call is audit-logged for your admins.
https://mcp.oystercatcher.ai/mcpStreamable HTTP · OAuth 2.1 · included on every paid plan and the 14-day trial
- Open Settings → Connectors and choose Add custom connector.
- Paste the server URL.
- Sign in to Oystercatcher on the consent screen. Done.
- Turn on developer mode (Settings → Connectors → Advanced).
- Add a connector with the server URL.
- Sign in to Oystercatcher when prompted and approve.
- Claude Code:
claude mcp add --transport http oystercatcher https://mcp.oystercatcher.ai/mcp - Cursor: add an MCP server, transport streamable HTTP, same URL.
- Complete the browser sign-in on first use.
Salesforce, Gmail or Outlook, and Apollo or Clay connect the same way, through each assistant’s own connector directory. Once all four are in, the Monday-morning prompt above runs as written. Full setup detail lives on the connector page.
Give one rep the prompt. See what Monday looks like.
Want proof first? Read an example dossier — sample data, real mechanics. See pricing for quotas.