I automate the important but mundane tasks inside a business with AI.
I work with my AI to take on jobs you want done. It can be a one-off task, or I set up an automation to do the same task repetitively. Most jobs are done in a couple hours.
The jobs that I've recently done: large-scale messy data migration between systems, mining tacit knowledge from email archives, anonymizing CVs and transcripts with an edge AI that lives on your laptop, custom visual AI systems for automatic architectural takeoffs of residential plans, automatic translation, reformatting, and restructuring of training material, high-quality automatic bilingual website redesign and rebuild, inter-Claude knowledge management and messaging system with mirror agents, AI-powered CRM and accounting system for small organizations, and much more.
Each of these started as a real problem for a real business (sometimes my own). Click any card for the full story.
A British e-commerce business needed years and years of product data migrated between systems — over 6,000 SKUs, each with a multitude of options. Their internal estimate was about a month. We did it together over two Google Meets, in about three hours.
Read the story →
For my other business, CoachingOurselves, I analyzed 18 years of my own sales correspondence — 438,000 messages I wrote — and extracted 736 high-signal customer recommendations. That dataset now feeds a pattern library my sales and marketing partner's AI assistant draws on every time a customer asks us which topics to run.
Read the story →
Recruiting agencies share candidate CVs with hiring managers every day, and under Quebec's Law 25 that means personal data has to be handled carefully. I built a tool that takes an emailed resume, strips names, contact details, and employer identity on a Montreal server, and only then hands the scrubbed skills and experience to the cloud AI that writes the summary and formats it into the recruiting agency's own template before it goes to the hiring manager. The raw resume never crosses the border. I built it for fun — it is the proof of concept for an anonymization method of my own that I now use on every project that touches private data.
Read the story →
CoachingOurselves publishes peer-learning modules. A CHRO forum asked for a 35-minute version of one of them. I wrote the cutting rules down first, then wrote a script that applies them to the InDesign source — deletes the cut pages, rewrites the elapsed-and-remaining counters on every kept page, redraws the progress bars, and rebuilds the agenda. Two production runs, same protocol, two very different pipelines.
Read the story →
I rebuild small-business websites using AI, with a fellow AI geek from Applied R&D. We scan the existing site, draft a new version in English and Quebec French, put the draft through a layered quality-assurance pass that catches typography, tone, and French rendering bugs, and deliver a complete bilingual redesign in a fraction of the usual time.
Read the story →
Academic researchers and consulting firms hit the same wall: before an interview transcript can be shared, every name and identifying detail has to be scrubbed and consistently replaced across the whole batch. I built a Windows desktop tool that does it in seconds — nothing ever leaves your computer, and the same person gets the same pseudonym in every transcript she appears in.
Read the story →
I run AI enablement workshops, and I always wanted to know what the people in the room needed from AI before I walked in. So I built an agent they email a few days before the workshop — it asks what they're trying to get AI to help with, learns what each person is wrestling with, and notices when two people describe the same problem in different words. By workshop day the agenda has already written itself out of named pain points the team already has.
Read the story →
A residential installer who does exterior windows and doors needs something they can drag an architect's PDF into and get back an accurate spreadsheet takeoff — every window, patio door, door and garage door — in minutes instead of an hour of manual counting. (In active development.)
Read the story →Most engagements start the same way. Someone introduces us. We get on a call and look at one thing you actually want done — something concrete, not a hypothetical. We spend about an hour together, live, on your real task, and at the end there is usually something tangible in your hands — a cleaned-up dataset or a working prototype.
If the work continues from there, we agree on a rate and carry on. I've built up a library of protocols for the kinds of jobs that come up repeatedly, which is why a short session is usually enough to produce something useful.
I'm Philip LeNir. I spent 15 years as a software engineer at SpeechWorks and Nuance Communications, building speech recognition and speaker verification systems — I hold a patent in speaker verification and did a lot of systems engineering along the way. For the last 18 years I've been running CoachingOurselves with Henry Mintzberg, a peer-learning business that taught me how real organizations actually work, what middle managers are up against, and how leadership and learning play out day to day. Senior engineering and senior management experience, combined. In late 2025 I started using Claude full time and found I could do things in hours that used to take me weeks, so I decided to build my consulting practice around that.
I'm based in Montreal and I work in English and Quebec French. Because I've been working with AI here in Quebec, I've become very good at working with Law 25 and the specific issues it presents for Quebec businesses and organizations.