Theodoros Ampas
Join the newsletter

EMBEDDED AI PARTNER

I build AI systems that run sales and marketing workflows

I work inside your team to turn repetitive sales and marketing work into systems your people can run.

Join the newsletterStart a project
A Mediterranean workspace with a laptop, system diagrams, and planning notes
Owned by you, not rentedEmbedded AIHuman reviewClear handover
Hands arranging workflow notes on a sunlit table

Forward-deployed. Not an agency.

I work beside the people doing the job, build AI agents and automations into the tools you already use, and stay until the system can operate without me.

  • A capable person is carrying repeatable work that should be a system.
  • Context is scattered across inboxes, documents, CRM records, and people.
  • AI experiments exist, but nobody owns a dependable production workflow.
  • The company wants capability it owns, not another rented dependency.

Real systems, built around real operating problems.

DMC lead loop with write-back

Problem
Incoming travel enquiries need qualification, follow-up, and a reliable record in the CRM.
System
A lead loop that prepares the next action, keeps human approval, and writes the accepted update back to the CRM.
Outcome
Baseline comparison in progress. The verified operational result will be published when the review is complete.
See AI workflow automation

SimplySites section-level tracking

Problem
Page-level analytics cannot show which sections help, stall, or lose a visitor.
System
Behavioral tracking tied to individual page sections, with events the product team can inspect in its own analytics stack.
Outcome
Product-decision and conversion impact are being verified before publication.

The Deal Database that scans Greece

Problem
Used-car listings are fragmented, while US valuation shortcuts do not represent the Greek market.
System
A local listings database, Greek comparable analysis, and a Telegram review queue with manual seller approval.
Outcome
Time, coverage, and deal-quality results are being verified before publication.
Read the Deal Database teardown

Embedded while it matters. Owned when it works.

Find the constraint

Map the repeated work, the decisions around it, and a baseline the team can verify.

Build in place

Connect the system to the repository, tools, permissions, and data the company already controls.

Make improvement supervised

Add feedback, memory, evaluation, monitoring, and explicit human checkpoints.

Transfer ownership

Document, train, stabilize, and leave named people able to operate and review it.

Practical notes from the builds

Read all articles
“AI works when your team stays in control.”
Theodoros Ampas

Learn how these systems are built.

The newsletter shares practical breakdowns first. If you already know the workflow that hurts, start a project.

Join the newsletterStart a project