Forward-deployed AI engineering for business teams
You describe the outcome. We ship the system.
A senior AI engineer embedded with your strategy, marketing, or research team, turning intent that is not yet a specification into systems running inside your stack. The industry standardized the name in 2025: forward-deployed engineering. We have been doing the work since 2014.
Delivered with IT kept close: their governance, their stack, their sign-off.
You should not have to write technical requirements
Seventeen years ago a marketing director told us the truest thing anyone has said about requirements: "Show me what you have and I'll tell you what I want." That is still how real requirements surface on the business side of the house. So we build the artifact that does the eliciting: a working prototype your team reacts to, instead of a specification no one could have written up front.
The arc of an engagement
It starts with a free 30-minute scoping call: no pitch, a straight answer on whether this fits. From there, four rungs, each one earning the next, with fees crediting forward as you climb.
1. AI Use Case Triage
Building the wrong thing correctly is expensive, and AI makes that mistake cheaper to make and harder to see. We name what not to build as clearly as what to build.
Bring your list of AI ideas. In 90 minutes, find out which to put first, which to stop, and what data questions need answers.
Within 48 hours, you get a two-page decision brief plus appendix: each idea graded PROCEED, CAVEAT or REFUSE (first pass), the data it touches, and the source behind every claim. You keep the brief whether or not you continue. The fee credits in full to the Discovery Sprint.
2. AI Discovery Sprint
We sit with the people who do the work: not a one-hour meeting, the actual work, and map what we see. You get an operating map of the workflow as it truly runs, and an ROI matrix: which workflows are worth automating, which are not, and what each is worth in revenue, cost, or risk.
Your Use Case Triage credits in full, and the sprint fee credits toward the prototype within 90 days.
3. Working Prototype
See your priority workflow working in your own stack. We agree what the prototype must demonstrate, then show working versions early and refine them with your team. You leave with a tested prototype, a clear account of its limits, and the scope for a production decision.
Project already underway but stalled? Same rung: we diagnose why the demo never reached production, then ship it or recommend killing it, with a clear alternative.
The delivery window starts on an agreed date, with data access, the working environment and review sessions in place.
Haven't done a Discovery Sprint? It's included. Already completed one? Your fee is credited.
4. Forward Deployed AI Engineering
A senior AI engineer embedded with your business, not beside it: shadow mode first, autonomy as the evidence earns it, your people leveling up while real work ships.
A forward-deployed engineer with no product to deploy: nothing to lock you into, we extend the stack you already run. Each quarter you get the numbers your board reads: hours saved, tasks shipped, risks retired.
Enterprise programs (multiple teams, formal security review, procurement) are scoped and quoted per RFP, and typically run into six figures. The rungs above are sized for a single team.
Why this works
Built to make its champion promotable
Bringing in outside AI help is a risk for whoever signs off. We know, because we have been the bet someone made. So the engagement is designed to de-risk YOU: visible milestones, deterministic checks on anything that must be right 100% of the time, an audit trail for every agent action, and a quarterly report with the numbers your board reads: hours saved, tasks shipped, risks retired.
Book a scoping callWho we work with
Strategy, Marketing and Research leaders
You can see what AI should do for your function. What you cannot get is engineering that understands you. That translation gap is our specialty: senior technical depth with an ear for the business, since 2014.
Companies where the business feels unheard by IT
When every request becomes a ticket and every ticket becomes a quarter, momentum dies. We work for the business side, with IT kept close: their governance, their stack, their sign-off - your velocity.
For consulting and professional-services firms
You win the strategy work; don't watch a vendor capture the build. Magma adds forward-deployed AI delivery to your bench without you building the practice: co-delivery or fully behind the scenes, under your paper, your NDAs, your brand. Your client is never our lead: no direct sales, no poaching, in writing. It is how we already work: our enterprise delivery history runs through consulting partners.
The role, before it had a name
"I'd like to highlight Daniel's 'superpower' - his ability to bridge the conversations between the 'business' and 'IT'. In fact I described him then as a 'unicorn' and he remains one of the few IT professionals I've worked with who have an uncanny ability to translate business needs, shield us from the complexity of the back end and work to provide solutions that benefit all stakeholders. His natural curiosity, excellent communication skills and solutions orientation sets him apart."
Nicole FrancisDirector, Sales Enablement, Softchoice
Delivered through Google Cloud PSO, Quantiphi, Intuitive.ai, and Ciklum. Led by Daniel Zivkovic, delivering production systems since 1996.
Google Professional Services caliber, Canadian pricing.
Results that speak for themselves
Frequently asked questions
How much does this cost?+
Our entry point is the AI Use Case Triage at $1,500: a 90-minute working session with your budget owner in the room, and a two-page decision brief plus appendix within 48 hours. The two-week AI Discovery Sprint starts at $15,000, scoped to your organization, and every dollar of the triage credits into it. Prototype and embedded pricing is scoped in a conversation, fixed before any work begins, with the sprint fee credited within 90 days. Enterprise programs (multiple teams, formal security review, procurement) are scoped and quoted per RFP, and typically run into six figures. The numbers here are the same ones we would quote on a call.
Why start with 90 minutes instead of a bigger assessment?+
Because the biggest mistakes are made at the start, when you choose what to build. A small error in direction there sends the project far off course later, and good engineering alone cannot fix choosing the wrong thing: you end up building the wrong thing correctly. So I spend 90 minutes getting the first decision right with the person who owns the budget, then put two weeks of Discovery on the one idea that earned it.
What happens in the Triage, and what does the decision brief cover?+
The budget owner joins the session, and you complete a short intake listing your AI ideas 48 hours ahead. Bring up to eight ideas. Three get examined in depth in the session; the rest get a first-pass grade and the question each still needs. There we agree the list, narrow it together, and examine the priority cases one at a time: who needs what and why; whether the system would read, write or act; what your existing vendors already cover. The two-page decision brief plus appendix is the decision record: which idea to put first for Discovery, what to stop and why, the data each idea would touch, and the question to answer first on each, with the person who owns the answer. Cases we did not reach get CAVEAT with the open question stated. You keep the brief whether or not you continue. Mapping the real workflow and what it is worth needs observation, so that belongs to the Discovery Sprint. See a sample report →
How long does this really take?+
Discovery takes two weeks once the workflow observation sessions and materials are arranged. The prototype then has a planned window of six to eight weeks, starting on an agreed date with usable data access, an approved environment and reviewers booked. Allow eight to ten weeks for both stages, ending at a tested prototype, not production adoption, plus any gap needed to secure access or a delivery slot. You see working versions during delivery, and the proposal sets the first demonstration date so you can judge progress before handover.
What happens if access or approvals take longer?+
We confirm the data, permissions and people needed before agreeing the start date. If a dependency later blocks progress, we continue any work that can move forward and agree revised dates for the rest. Waiting alone does not increase the fixed fee. A long pause may need a new delivery slot, and any extra work is scoped and priced for your approval before it begins.
Will our IT team be okay with this?+
It is designed for that. We work inside IT's governance: their stack, their security review, their sign-off. Nothing ships around them, and autonomy ramps in shadow mode until the evidence earns it. We arrive with an audit trail, not a shadow system: IT gets a governed system to review, not one to unwind.
Isn't "forward-deployed engineer" a Palantir term?+
Yes: Palantir coined the role two decades ago, and in 2025 the industry adopted the name broadly. The job itself, a senior engineer embedded with the business side, turning intent into running systems, is what Daniel has done since 2014, before the label traveled. Our field guide to the model tells the full story.
We are a consulting firm. Can you deliver under our brand?+
Yes, that is how most of our enterprise work has always shipped: co-delivery or fully behind the scenes, under your paper and your NDAs. Your client is never our lead: no direct sales, no poaching, in writing.
How is this different from hiring a consulting firm?+
You get the senior engineer who built systems for the logos above, directly. No B-team, no markup for project-management overhead. You pay for talent, not org chart.
Can one person really handle enterprise projects?+
I cover the full SDLC: not just implementation, but the discovery, architecture, security, and governance work that gets skipped when teams panic. For larger engagements, I partner with Google Cloud PSO and specialist firms. The work behind those logos was production delivery at Fortune 500 scale.
Do you only build, or strategy too?+
Both. Alongside the delivery work, Magma offers Strategic Foresight: a cited, dated, six-lens read on the shifts coming at your market, so your team can adjust positioning and place its bets early. Same evidence-graded method, applied to your market instead of your codebase.
What if I just want to talk first?+
Book a 30-minute scoping call. No pitch. Bring your messiest AI question and I'll give you a straight answer. If there's a fit, I'll tell you. If there isn't, I'll point you in the right direction.
Start the conversation.
A 30-minute scoping call: no deck, no pitch, a working conversation. If a sprint fits, your operating map lands within two weeks of starting.
No pitch. Bring your messiest AI question. I'll give you a straight answer.
Want to read before you talk?
A field guide and the other half of the job. Start with whichever question is more pressing.