OrcaScale
orcascale.co / agents
RESEARCH · SCORING · DRAFTS · APPROVALS
The flagship · In service today

The same agentic employee that runs OrcaScale's own outreach

This is not a concept card. This exact pipeline does OrcaScale's own prospecting, end to end, every week, and we can show it to you working. Everything else in the catalogue is built with the same rules: real tools, real evidence, and a human at the gate.

REF 01 / 06 · Outreach ResearcherStatus: in service, running OrcaScale's outreach

Outreach Researcher

End-to-end prospecting pipeline

Its working day, start to finish: it picks a city and an industry, finds real operating businesses there, and researches each one against its actual website in a rendered browser. It scores the evidence, rejects weak prospects outright, writes a personalised pitch for the ones that survive, puts that pitch through an adversarial QA review, and leaves a finished draft in Gmail.

  1. Picks a city and an industry to work
  2. Finds real operating businesses there
  3. Researches each one against its live website
  4. Scores the evidence and rejects weak prospects
  5. Writes a personalised pitch for the survivors
  6. Runs the pitch through an adversarial QA review
  7. Leaves a finished draft in Gmail for a human
28Prospects worked in its latest run
14Rejected for weak evidence or no contact route
0Emails the agent is able to send

The rejection rate is a feature, not a failure. An agent that comes back and says "no good prospect here" is worth more than one that always finds something, because you can trust the ones it does put forward.

Works inGoogle Sheets · Gmail · a rendered browser
Approval gate

A human reads and sends every email. The agent cannot send; the most it can ever do is leave a draft in Gmail.

The catalogue

Five more roles, built to order

Each profile is a role Ed builds for businesses like yours, working in your existing tools or inside a custom OrcaScale system. This is a catalogue of what gets built, not a claim that an army of them is already deployed everywhere. Read each one like a personnel record: what it does all day, where it works, and exactly where you stay in charge.

REF 02 / 06Status: built to order

Industry & City Researcher

Market mapping, verified at source

Give it a market and a place, and it builds a verified list of real businesses that operate there, with what each one actually does in its own words, taken from its own website rather than a stale directory. Defunct listings, lookalikes and scraped junk get filtered out before you ever see them.

Works inGoogle Sheets · your CRM
Approval gate

You review the list before anything downstream uses it. Every entry links back to the source it was verified against.

REF 03 / 06Status: built to order

Lead Scorer

Evidence-ranked pipeline

Reads your inbound enquiries or a prospect list and ranks every lead by evidence, not hunches. Its reasoning is written down next to every score, so you can see exactly why a lead ranked where it did, and overrule it in seconds.

Works inGoogle Sheets · your CRM · your OrcaScale system
Approval gate

Scores are recommendations. A human sets the cut-off and decides who gets pursued.

REF 04 / 06Status: built to order

Personalisation Writer

One line that proves you looked

Turns research into the one specific line that proves someone actually looked, never a template blast. It works only from verified facts about the prospect, and it flags itself when the evidence is too thin to say anything honest.

Works inGmail · Google Sheets · your CRM
Approval gate

A human approves every message before it goes anywhere near a prospect.

REF 05 / 06Status: built to order

Email Drafter

Your voice, as drafts only

Drafts replies and follow-ups in your voice, learned from your actual thread history rather than a generic tone setting. It keeps the whole conversation in view, so the third follow-up still reads like you wrote it, not like a mail merge.

Works inGmail · your OrcaScale system
Approval gate

Everything lands as a draft in your inbox. Nothing sends itself, ever.

REF 06 / 06Status: built to order

Spreadsheet Keeper

The sheet that stays true

Keeps a working sheet honest. It pulls updates from the tools where the work actually happens, flags stale rows and inconsistencies, and writes a note against every change, so the sheet stops quietly drifting away from reality.

Works inGoogle Sheets · your CRM · your OrcaScale system
Approval gate

It never overwrites without a trail. Every change is logged, attributed and reversible.

Do not see the job you would hand over? These six are the roles we get asked about most, not the limit. If the job repeats, lives in real tools and can be checked before it matters, it can probably be an agent.

The honest bit

Is this not just ChatGPT?

Fair question. A chat window is genuinely useful, and we use one every day. But there is a real difference between a tool that answers questions and an employee that owns a job.

A chat window

Answers when asked

  • Answers questions when you ask, then stops until you come back
  • Lives in a browser tab, apart from the tools where the work happens
  • Starts every conversation from a blank page
  • Leaves nothing behind that anyone can audit
An agentic employee

Owns a repeating job

  • Owns one repeating job and runs it on a schedule or a trigger
  • Works inside your actual tools: Gmail, Google Sheets, your CRM, your system
  • Shows its work: what it looked at, what it decided, and why
  • Stops at a named approval gate, where a human has the final say
And the honest limit: agents are wrong sometimes. Not often, but sometimes, and confidently. That is exactly why every agent here ships with an approval gate and an audit trail rather than free rein. You get the hours back without handing over the judgement.
Yours, not rented

It runs in your accounts, and you own it

Every agent we build runs in your accounts: your Gmail, your Google Sheets, your CRM, your OrcaScale system if you have one. There is no third-party platform sitting between you and your own data, and nothing to migrate off if we ever part ways.

An optional monthly care plan keeps it sharp: you run the business, we keep the agent honest. And it is optional either way. You own the code, the prompts and the data, whether you take a care plan or not. Full details are in the FAQ.

  • Runs under your accounts and credentials
  • You own the code, the prompts and the data
  • Approval gates and audit trail on by default
  • Monitoring, updates and tuning under the care plan
  • Priority support when something matters
How it starts

Tell us the job you would hand to an agent

About ten minutes at /builder: the repeating job, the tools it lives in, and where you want the approval gate. Ed reads every application himself, and if a plain script or an off-the-shelf tool is the honest answer, he will say so.

One repeating job, handed over. Your tools, your approval gate, your agent. Apply for a free build See all case studies