Test tag · illustrativeOrdering agent
- Condition
- A customer asks to change the address on an order that has shipped.
- Must not
- Change a dispatched order on its own.
- Who's told
- A person, with the order and the request attached.
Services
Built around what's actually stuck.
Every system is custom and yours to keep, against a yearly fee that covers keeping it running. Each one is connected to the tools you already use, knows when to fetch a person, and gets tested before it ships.
01 of 03 · E-commerce systems
Your store, watched by agents that write to one record.
We start with the order: WhatsApp ordering that confirms cash on delivery and checks stock before anything ships, returns handled the same way, and anything unusual handed to a person.
For a whole store, we also built πCounter: “Five agents watch your storefront, orders, customer conversations, ads and money, write to one record, and bring you the change with the evidence attached.” It lives at picounterai.com.
- Message in
- Intent read
- Stock checked
- Order placed
- Person told
Works with Shopify · WhatsApp · Razorpay · Stripe · Instagram
02 of 03 · Sales & GTM systems
Inbound calls answered, even at 11pm.
Each caller is qualified against what you're actually looking for, and the ones worth your time get booked. Outbound lists are enriched before anyone picks up the phone. You get a summary, with the transcript there if you want it.
Composite example · solar installer, 14 people
Four calls came in after 6pm on Friday. Two left voicemails. By Monday, both had booked with someone else.
Now every call is picked up, qualified against what they're looking for, and booked, with a summary waiting for the owner.
- Call in
- Intent read
- Lead qualified
- Meeting booked
- Person told
Works with HubSpot · Apollo · Clay · ElevenLabs · LinkedIn · WhatsApp
Test tag · illustrativeCalling agent
- Condition
- A caller asks for a price the script can't quote.
- Must not
- Guess a number to keep the call going.
- Who's told
- The owner, with the question and a call-back slot.
03 of 03 · Recruitment systems
No application sits unread.
It reads every résumé that comes in, the last as carefully as the first, and brings forward the candidates worth a real conversation. People make the decisions.
- Résumé in
- Screened against your criteria
- Shortlist surfaced
- Person decides
Works with LinkedIn · Gmail · Google Sheets · Slack
Test tag · illustrativeScreening agent
- Condition
- A strong application that misses one stated requirement.
- Must not
- Reject a person on its own.
- Who's told
- Your hiring lead, with the reason it was flagged.
Runs through all of them · AI Safety Audit
Whatever you already have running, we spend two weeks trying to break it.
Where your AI leaks data, gets manipulated, or fails quietly. You get a ranked risk report and a fix order you can act on the same week. Two weeks, fixed scope, priced as a project.
- System mapped
- Attacks attempted
- Failures logged
- Report delivered
Always included
Audit, build, train, stay.
- An audit before anything reaches a customer
- Custom code with error handling from day one
- Training so your team can run it and change it
- Two weeks of live monitoring after launch, then we step back.
You own it: the repository, the documentation and deployment access. It runs on infrastructure we manage, and one yearly fee covers maintenance, hosting and the AI usage behind it. Stop paying it and you keep the code, but it stops running until you host it somewhere else.
Also built on request
The same care, for other stuck work.
Website chatbots
answer from your own pages, refuse what they shouldn't, and log every attempt to trick them.
Content engines
brand documents, topics, drafts, visuals and publishing, for one brand or several.
Operations desks
leads, stock and the team's daily brief on one board, with an agent sorting what came in.
Claude Code and Grok, set up for your team
connected to your tools, with the access rules written down.
Questions we get
Before you ask.
How long is the first call?
30 minutes, free. You'll know by the end whether an agent is the fix.
Which tools can you connect?
Anything with an API, and tools that support MCP. The ones we use most: Shopify, WhatsApp, HubSpot, Apollo, Clay, ElevenLabs, Google Sheets and Slack.
How long does a build take?
Two to six weeks after a one-week design phase, then two weeks of live monitoring.
Do we own what you build?
Yes. The repository, documentation and deployment access are yours, and we train your team to run it. It runs on infrastructure we manage, and one yearly fee covers maintenance, hosting and the AI usage behind it. Stop paying it and you keep the code, but it stops running until you host it somewhere else. Nothing takes the work back off you.
Which time zones do you work in?
UK, India, EU and Australian hours.
What if AI isn't the right fix?
We say so on the first call and point you at whatever is.