Five levels of Forward Deployed AI, from Use to Operate
Forward Deployed AI has five levels: 1 Use, 2 Connect, 3 Build, 4 Ship and 5 Operate. You pay a monthly retainer for engineers at the level the work needs, from level 1, Use, to level 5, Operate, and change the mix as the work changes.
The five levels side by side
| Level | What staff can do | What the engineer does |
|---|---|---|
| 1 Use | Write clear prompts, check the output, share what works | Licences, prompts, simple automations |
| 2 Connect | Link the assistant to company files, mail and systems | Connects tools, runs adoption |
| 3 Build | Build no-code tools and simple agents: AI that does one job by itself, checked by a person | Builds simple tools and agents |
| 4 Ship | Build and release software with AI coding tools, tests and access controls | Senior engineer, owns quality |
| 5 Operate | Run AI agents on core systems such as finance, with oversight | AI operating partner: builds the first prototype, advises the board |
Choose a level

Level 1, Use
At level 1, Use, a Forward Deployed AI engineer handles licences, prompts and simple automations. Your people learn to write clear prompts, check the output and share what works.
Entry price on the level 1 page

Level 2, Connect
At level 2, Connect, the engineer connects tools and runs adoption. Your people can link the assistant to company files, mail and systems.
Priced on the work

Level 3, Build
At level 3, Build, the engineer builds simple tools and agents. Your people can build no-code tools and simple agents: AI that does one job by itself, checked by a person.
Priced on the work

Level 4, Ship
At level 4, Ship, a senior engineer owns quality. Your people can build and release software with AI coding tools, tests and access controls.
Priced on the work

Level 5, Operate
At level 5, Operate, an AI operating partner builds the first prototype and advises the board. Your people can run AI agents on core systems such as finance, with oversight.
Priced on the work
Which level to start at
- 1 UseStart here if you have bought Claude, ChatGPT or Copilot and few people use it every day.
- 2 ConnectStart here if people use the assistant but it cannot reach your files, mail or systems.
- 3 BuildStart here if a team has a repeated job that a simple tool or agent could take on.
- 4 ShipStart here if a pilot works and now has to run in production, tested and owned.
- 5 OperateStart here if the board wants AI agents running on core systems, with oversight.
A team below level 1 starts with a level 1 engineer.
How an engagement runs and grows
Before week 1
One goal agreed with you for the six weeks: a measure, a baseline and a target.
Week 6
The goal is reviewed, and you decide whether to continue and what the next goal is.
Three months
The minimum term, covering two six-week goals.
After that
Month to month. Either side can give one month's notice.
A typical path starts with one level 1 engineer. After one or two sprints, when the next opportunity on the list is a single-task agent, the retainer moves to level 3. When that agent needs to run in production, a level 4 senior engineer joins to own its quality. Some clients add a level 5 AI operating partner for the board conversation while the level 1 engineer keeps adoption moving.
Higher levels and further engineers are priced on the work. When the work outgrows one team, a larger build goes to QuantSpark Labs: larger builds, platforms built and supported for you.
Forward Deployed AI is the Pilot stage of QuantSpark's four-stage model
- 01
Prepare
Get ready for AI
- Data readiness: cleansing and readiness
- Workflow mapping
- Training and people readiness: adoption and skills
- 02
Plan
AI advisory
- AiRE, our AI Rollout Engine
- An AI vulnerability and opportunity assessment
- An AI roadmap and buy-versus-build
- AI governance and security posture
- 03You are here
Pilot
Forward-deployed engineers
- Forward-deployed engineers and consultants work embedded in your team
- Agents, skills and pilot products are built and tested
- 04
Productionise
Enterprise solutions
- Full builds through QuantSpark Labs
- Integrate, measure and scale
- Support and maintain developed solutions
Clients can enter at any stage; what we learn in production sharpens the data, plans and pilots that follow.
Questions about the levels
What are the five levels?
The five levels are 1 Use, 2 Connect, 3 Build, 4 Ship and 5 Operate. Level 1 covers licences, prompts and simple automations. Level 2 connects tools and runs adoption. Level 3 builds simple tools and agents. Level 4 is a senior engineer who owns quality. Level 5 is an AI operating partner who builds the first prototype and advises the board.
Which level should we start at?
Start from where your team is today. A team below level 1, where people are not yet using the tools every day, starts with a level 1 engineer. If the assistant already needs to read company files, mail and systems, or adoption has stalled in some teams, level 2 fits.
Can we move up a level later?
Yes. You pay a monthly retainer for engineers at the level the work needs, and you change the mix as the work changes. A typical path starts with one level 1 engineer. After one or two sprints, when the next opportunity on the list is a single-task agent, the retainer moves to level 3. When that agent needs to run in production, a level 4 senior engineer joins to own its quality. Some clients add a level 5 AI operating partner for the board conversation while the level 1 engineer keeps adoption moving.
Do you work with Claude, ChatGPT or Copilot?
Yes, as the tools you roll out. Our engineers are model-agnostic. They build on the models and agent platforms you already use, from Anthropic, OpenAI, Google or open models, and connect your tools through the Model Context Protocol, the open standard for linking AI to company systems. Claude, ChatGPT and Microsoft Copilot are assistants companies roll out, and the engineer works with whichever you choose.