Print on demand ~20% → ~40%
conversion to a final sale
For mid-market companies
End-to-end agentic AI transformation, for leaders who want it shipped.
Boutique partnership. Senior-led. Fixed scope. A limited number of engagements per year — one operator accountable end-to-end, with the Vstorm engineering team shipping the build.
The challenge
Three reasons people usually book the first call.
Each is a different project shape. All three need someone who has already done this work for a company your size.
Context A
Transform your company processes with agents.
You already have operations that work — but they're expensive, slow, or depend on people doing repetitive judgment work. You want to re-engineer them with agentic AI.
Context B
Build an agentic product as your new core service.
Your service — the thing you sell — can itself be delivered by agents. That is a product and business-model change, not a bolt-on feature.
Context C
Something else entirely.
Exploring a thesis, pressure-tested by your board, sorting signal from noise. Book a call and we will figure out if it is worth a conversation.
Why this is hard right now
Everyone talks about agentic AI. Almost nobody ships it.
01
The vocabulary is ahead of the operations.
Most board-level conversations sit one layer of abstraction above anything that can be built. The words exist; the operating knowledge does not.
02
Most agentic AI initiatives stall.
The failure mode is almost never the model. It is the surrounding system — data, workflows, evaluation, change management.
03
The project shape has to change.
Agentic systems are probabilistic, they interact with humans, and they fail in ways traditional QA does not catch.
04
Few service firms have production implementations.
Big-4 packages often stop at the deck. Local agencies reskin chatbots. Very few teams have put agents into production for mid-market clients.
The cost of waiting
Four things compound against you. All four are already in motion.
Mid-market feels this first. Enterprise can throw budget at delay; startups lack the surface area. Twelve months of indecision changes the shape of your P&L.
01 Competition
Competitors put agents into workflows you still run by hand.
Their unit economics move 6–12 months before yours. By the time it shows up in a market report, you are already two quarters behind on margin and win rate.
02 Talent
Your best operators leave.
Strong people want modern tooling and problems where agents take the repetitive judgment work. Teams that do not offer that bleed quietly — usually the top quartile first.
03 Strategy cycle
Next year becomes another "AI strategy" engagement.
Then another. Same deck, four times. A year of compounding gone, still no agent in production. The deck was never the hard part.
04 Market pricing
Agentic AI talent gets more expensive every quarter.
Entry price keeps rising — teams, tooling, and partners cost more the longer you wait. Starting late costs time and a different budget envelope.
The work
End-to-end transformation — from the boardroom to the build floor.
I do not stop at a deck, and I do not start at the IDE. I run the whole ladder: consulting, opportunity identification, planning, deep engineering, and — when the company is ready — an AgenticOS layer that wires agents into how the business actually runs.
One operator. The whole ladder.
Most firms sell a slice — strategy, or PoC, or engineering. The slices do not compose. You end up owning the integration risk yourself.
I run each engagement personally. The team at Vstorm delivers the build. Thesis, ship, and handover stay with the same operator.
How we build at Vstorm →
01 CONSULT
Transformation consulting
Board-level framing. Where agents change the business, and where they don't. A thesis your sponsor can defend.
02 DISCOVER
Identifying opportunities
Process mapping, data readiness, integration surfaces. A ranked list of high-impact agent interventions.
03 PLAN
Planning & architecture
Roadmap, eval strategy, guardrails, and org design. The plan is small on purpose — one agent, real users, six weeks.
04 BUILD
Deep engineering
Senior engineers, in production. Tooling, observability, and evaluation loops built as first-class citizens, not afterthoughts.
05 OPERATE
The full AgenticOS
Agents across finance, ops, sales, support — wired into one operating layer. Humans keep judgment, taste, and relationships; agents take the repetitive work.
Board deck Small giant
The outcome
A small giant, built on an AgenticOS.
Boutique, senior-led, disproportionately profitable — mid-market scale with the agility of a small team.
Agents handle the repetitive work. The company still needs people for judgment, taste, and relationships — with a durable agent layer underneath.
Inspired by Burlingham's Small Giants, built for operators shipping now
How we work together
Three stages. We call it TriStorm.
We start with the process you run today and finish with a system your own team runs and owns. No stage starts before the previous one is signed off.
- Stage 01
Process redesign
We map the process you run now and redesign it for agentic AI, so the system can run it more efficiently than people do.
- Stage 02
Agentic process transformation
We engineer the agentic AI system that transforms the process.
- Stage 03
Hand-over, team upskilling
We hand the process over turnkey, upskill the team that works with it, and build the capability inside your organization. You own the system fully, build AI sovereignty, and do not rely on any vendor in the future.
Why me
Four reasons that hold at once.
01
I have shipped.
Agents in production for mid-market clients across Europe, US, Canada, Saudi Arabia. References available on request.
02
Senior-led, end-to-end.
I run the engagement myself. One phone number, one operator accountable for the life of the project. Vstorm engineers the build.
03
Operator and investor.
I back B2B agentic AI startups with personal capital — so I see tooling and patterns months before they are common.
04
You can watch me think.
Read the essays on this site. Decide whether I understand the work before you book.
Proof
Shipped, not pitched.
Short summaries of engagements I ran. Long-form case studies live on vstorm.co — the delivery company that ships the builds.
30+
Production deployments Through Vstorm, in production.
2017
Shipping AI since Vstorm founded; still running it.
Europe, US, Canada, Saudi Arabia
Delivery footprint Named markets on the company record.
Print on demand 
Engineering software 2 hrs → 3 min
to generate a validated workflow
Clinical triage 44% → 98%
accuracy on an open 50-scenario clinical triage benchmark
Questions
What leaders ask before the first call.
Q1 How long until the first agent is in production?
Six weeks from scope lock. Week one locks one workflow, one owner, one system on a single page. Weeks two to five build the agent on real data, then harden it with evaluation loops, monitoring, and real users. Week six is documentation and handover.
Q2 What does it cost to start?
The first call is 30 minutes and costs nothing. Engagements run a minimum of 90 days with a realistic budget. I do not quote a number before discovery — scope decides it. For reference: the median case I see is an agent shipped in six weeks for a fraction of what a typical AI strategy document costs.
Q3 What happens if our data isn't ready?
That is the normal case. Data access and normalization take about 30% of engineering time on a six-week build — authentication, format normalization, and edge cases that miss the first ten test runs. The map stage covers data readiness before anything gets built. When I estimate data access time I triple it, then double it again.
Q4 Should we build this or buy it?
Usually both, at different layers. Build when the workflow is your competitive moat, when proprietary data is the core asset, or when the agent sits in your critical path. Buy when the workflow is horizontal and non-differentiating, when your team cannot absorb the learning curve, or when a competitive window closes faster than you can build. For most mid-market companies: buy the horizontal infrastructure and build the domain agents on top.
Q5 Who actually does the work?
I run the engagement myself. The senior team at Vstorm — the company I co-founded — delivers the build. Same person on the thesis, the ship, and the handover. One phone number for the life of the project. That is also why I take a limited number of engagements per year.
Q6 What does "shipped" mean here?
An agent running in production, on real data, with real users and real consequences — not a demo in staging. It has a named production owner inside your organization before the first line of code, plus evaluation loops, observability, guardrails and a human-in-the-loop fallback. Handover includes runbooks, alert thresholds and the edge cases we already know about. If a pilot has not named a production owner by week two, it is already a deck.
If this is the work you need