AI implementation

Run more with the team you have.

Boost throughput by clearing the manual computer work behind your bottleneck, with high-integrity AI.

Start a conversation See the work
How the work starts
Stage one

Prototype

We pick the one job costing you the most and build it, for a fixed price, in weeks. It runs on your own systems, and if you stop there it is yours to keep.

Stage two

Capacity

This is the real build: the connections between the systems you already run, and the tools your team works in every day. It only happens if stage one earned it.

Stage three

Compounding

Your systems keep changing and the models keep getting better, so we keep adding to what is already running. Each piece makes the next one quicker.

Sample projects
A room of executives working through a session together
01

Professional services

Automated client onboarding

Onboarding a new client used to mean retyping the same details into six systems before any work could start. It now happens on its own the moment a deal is marked won, so the folder, the account and the team channel are all ready before anyone opens a laptop. The welcome email is posted into that channel as a draft, because that one should still come from a person.

Live. The client has been running and extending it themselves for four months.

ValueTime savings
SectorProfessional services
StateLive
Ref01 / 05

Headless agent runtime, connector servers, event-triggered CI dispatch

Loaded pallet racking receding down a distribution warehouse
02

Equipment distribution

A single source for product data

The same product had different specifications depending on which file you opened, because the data was 39 GB spread across more than 14,000 files that disagreed with each other. They now have one catalogue behind the website, the spec sheets and the sales tools, with most of its values carrying a recorded origin and the hardest specifications carrying the quoted line from the manufacturer document they came from.

Live and feeding four systems. The verification queue is deployed and audited, and the client team has not yet started working its backlog through it.

ValueRisk and compliance
SectorEquipment distribution
StateLive
Ref02 / 05

Extraction and reconciliation pipeline, SQLite at the edge, entity-attribute-value schema, read-only connector

A product photography studio lit for a plain white background
03

Wholesale distribution

A complete online store

A distributor with hundreds of products had no way to sell them online. They now have a complete store: accurate specs on every product, real descriptions instead of thirty capital letters, consistent photography, and categories ordered so the things that actually sell appear first. Building it with scripts rather than by hand is what made that possible, and it meant every listing could be checked rather than taken on trust.

Built and fully populated. It has not been opened to the public yet, so there is no traffic or sales result to report.

ValueRevenue growth
SectorWholesale distribution
StateBuilt, not launched
Ref03 / 05

Commerce platform admin API, templated theme sections, background segmentation, vision and image-to-image models

The interior of a modern five-axis machining centre, spindle loaded over a trunnion table
04

Precision manufacturing

Part handling posted automatically

A machinist was pasting the same hundred lines into every program by hand, because the post processor could only add them at the end and the shop needs the grab and transfer partway through. The post now writes the eject and transfer itself, every job, with nothing typed in afterwards. Getting there meant reworking the post processor rather than papering over it in the G-code, and two faults turned up that the vendor still ships. One commands the sub spindle chuck to close in the line that says it opens.

In production at the client. The eject and transfer routine posts automatically now, nothing gets typed into the program by hand, and the shop maintains the file itself.

ValueTime savings
SectorPrecision manufacturing
StateIn production
Ref04 / 05

CAM post processor, G-code, machine builder reference tables, CAM toolpath templates, local connector on the client’s workstation

Two machined flanges resting on dimensioned engineering drawings
05

Precision manufacturing

Quote pricing from real costs

The owner priced every quote himself, from experience, with the drawings buried in email attachments. He now gets a suggested price on each request, anchored on what he charged for comparable parts, alongside the dimensions off the title block and what the metal actually cost on recent invoices. The CAD models open in the browser, so nobody has to launch CAD to spin a part during a quote.

Deployed and running on real data. It writes a draft estimate into the accounting system and a draft reply into his mailbox and stops there, because nothing should send without him. The end-to-end run-through with the owner has not happened yet.

ValueRevenue growth
SectorPrecision manufacturing
StateDeployed
Ref05 / 05

Edge functions and key-value store, frontier language model, delegated OAuth into live mail and accounting systems, in-browser CAD geometry viewer

What clients say

Derik came in to help us find automation opportunities in our leadership development agency. We work with a lot of client information across tools: sales calls, proposals, onboarding docs, project plans. My team was easily spending one to two hours every new client onboarding just moving information around and consolidating it.

He helped me map the flow visually on a whiteboard, then teaching me alongside him in Claude Code, we integrated all those tools into one process. The agent now pulls the necessary information from every section and brings it into one cohesive view. When a few things broke a few days after testing, he got on a call with me and we troubleshot it together.

Working with Derik to build these AI agents has been fun, very worthwhile, with a huge ROI for my team. I’d highly recommend it to anyone.

Fahd AlhattabFounder, Unicorn Labs

Derik at Thrive was instrumental in taking incredibly messy data we inherited in a business we acquired and, through using AI, organized it in record time in a way that made it reviewable by our team for final review and approval.

He proposed solutions that were effective to implement and everyone on our team was impressed and happy to work with him.

We’ve only started to work with him, hopefully many more projects to come with so many use cases for AI implementation in our businesses.

Rios-Karim MercierBelmont Capital
Contact