Two Models, Very Different Outcomes
The AI implementation market has split into two distinct models, and most business owners choose between them based on the wrong criteria.
Done-for-you: an agency builds your AI system. They scope the project, design the architecture, build the agents, and hand you a finished product. You pay for the output.
Done-with-you: a program guides you through building the system alongside an expert. You participate in the build, develop the capability, and own the knowledge of how the system works. You pay for the transformation.
Both are legitimate. Neither is universally right. The mistake is choosing based on convenience or budget when the real decision factor is something else entirely: the state of your IP.
The Hidden Problem With Done-for-You
Done-for-you AI builds sound efficient. You describe what you want, an expert team builds it, and you get a working system without the learning curve. For many businesses, this is exactly the right approach.
For founder-dependent $500k+ knowledge businesses, it usually is not. And the reason is structural, not about the quality of the agency.
An AI system that delivers your methodology needs to encode your methodology. Your diagnostic frameworks. Your feedback criteria. Your quality standards. The judgment calls you make at each decision point in your delivery process.
If that methodology is already documented, written down in enough detail that a new hire could deliver 80% of it from the documentation alone, then a done-for-you build has what it needs. The agency reads your documentation, translates it into AI logic, and builds a system that reflects your actual approach.
If that methodology is not documented, if it lives in your head, in your habits, in the intuitive pattern recognition you have built over years, then the agency is building on a gap. They will fill that gap with generic logic. The system will work. It will not work like you work.
14%
of business owners achieve consistent results when scaling. The IP required for consistency has usually not been extracted
Clients notice the difference. A diagnostic that asks generic questions instead of your diagnostic questions. Feedback that follows a reasonable structure but misses the specific criteria you apply. A system that produces plausible output rather than output that meets your standard.
The agency did their job. The IP was never properly transferred. The system reflects what the agency understood, not what you know.
An agency can only build what you can articulate. If your methodology lives in your head, the system will encode their interpretation. Not your expertise.
What Done-With-You Actually Involves
Done-with-you is not a polite way of saying "we make you do the work." It is a recognition that for founder-dependent businesses, the AI build and the IP extraction are the same process.
A properly structured done-with-you program has three phases:
Phase 1: Immersion. Build AI fluency from the ground up. LLM foundations, toolchain setup, industry landscape. You build something real — an interactive prototype — so you see what's possible. By the end, you think in AI.
Phase 2: Mapping. Extract your IP — every framework, SOP, judgement call. Name your methodology, design diagnostics, map constraints. One focused leverage plan. One prototype specification, build-ready. The founder understands what is being built and why.
Phase 3: Transformation. Build it with AI. Agents, skills, workflows — your IP encoded into a working system. Gates enforce quality. The system deploys live. Your team gets trained. When the engagement ends, the capability stays.
The output is not just a system. It is a team that understands their own methodology at a new depth and can deploy it through AI independently.
The Real Comparison
Done-for-You AI
Faster to deploy. Requires less founder time during the build. Works well when IP is already documented and requirements are clear. Creates dependency on the builder for changes. The system reflects what the agency understood, which may not be what you know.
Done-with-You AI
Slower initial build. Requires significant founder involvement. Forces IP extraction that has independent value. Builds internal capability. The system reflects your actual methodology because you helped define and test it. No ongoing dependency.
The comparison is not about quality of build. Good agencies build excellent systems. Good done-with-you programs produce excellent systems. The difference is what happens after the build.
With done-for-you, changes require the agency. New workflows require the agency. Modifications when your methodology evolves require the agency. Every iteration has a cost, a timeline, and a dependency.
With done-with-you, your team owns the system. They understand the IP layer, the agent logic, the quality criteria. When something needs to change, and it will because your methodology is not static, they can change it. The capability compounds internally rather than being rented externally.
When Done-for-You Is the Right Choice
Done-for-you is the right model when three conditions are met:
Your IP is documented. The methodology, the decision logic, the quality standards: all written down in enough detail that someone who does not know your business could follow them and produce output you would approve. If you can hand an agency a specification document and they can build from it without guessing, done-for-you works.
The requirements are stable. You know exactly what the system needs to do because you have been doing it manually for long enough that the process is settled. You are not experimenting with your methodology. You are scaling a proven one.
You have internal maintenance capability. Someone on your team can manage the system after the build. They understand the logic, can troubleshoot issues, and can make minor modifications without going back to the agency. The build is a one-time engagement, not an ongoing relationship.
If all three are true, done-for-you is faster and more efficient. The agency brings technical expertise. You bring documented IP. The combination produces a system that works.
If any of the three are not true, and for most founder-dependent businesses the first condition is the one that fails, done-with-you is the safer bet.
43.8%
of US businesses now pay for AI tools. Adoption without internal capability creates new dependencies
When Done-With-You Is the Right Choice
Done-with-you is the right model when the business is not ready for a pure build and recognises that readiness itself is the first deliverable.
This typically means:
Your IP needs extraction. The methodology is in your head. You can deliver it but you cannot hand it to someone else, whether a team member or an AI system, and get the same result. The done-with-you process forces the extraction that makes any AI build viable.
You need to understand the system. Not at a code level, but at a logic level. What the AI is doing, why it is doing it, and how to know when it is doing it well. If you cannot evaluate the output of your AI system against your own standard, you are not in control of your delivery quality.
You want to iterate independently. Your methodology evolves. Your market changes. Your program develops. The AI system needs to evolve with it. If every change requires an external engagement, iteration slows to the speed of procurement rather than the speed of insight.
The AI Transformation Readiness framework covers how to assess where your business sits on this spectrum and what needs to be true before any AI engagement produces lasting value.
The question is not who builds the system. The question is who understands it well enough to evolve it after the build.
AI Consulting vs AI Agency: The Same Decision in Different Words
If you have been weighing an AI consulting engagement against an AI agency, you are looking at this same choice from a different angle. An AI agency is the done-for-you model: they build and hand over. AI consulting, done properly, is the done-with-you model: an expert works alongside you, extracts your IP, and leaves you with both a system and the capability to run it.
The label matters less than the structure. Ask any provider one question: at the end of this, do I own a black box, or do I own a system my team understands? That answer tells you which model you are actually buying, whatever it is called.
For a fuller breakdown of how to tell genuine AI consulting apart from agency delivery, see how to choose an AI consultant and the comparison of AI consulting firms.
The Sequencing Mistake
Some founders try to shortcut the process: hire an agency for a quick done-for-you build, then learn the system afterwards. Build first, understand later.
This rarely works. The agency makes design decisions based on their interpretation of your methodology. Those decisions compound, each one shaping what the next one can be. By the time the system is handed over, the architecture reflects the agency's understanding, which may diverge from yours in ways that are difficult to unwind.
The better sequence for most founder-dependent businesses:
- Extract the IP (done-with-you or independently)
- Build the first AI system (done-with-you, so you understand the logic)
- Scale and refine (internally, with the capability you now have)
- Engage agencies for specific technical builds where requirements are clear (done-for-you, now that the foundation is solid)
Done-for-you is a legitimate tool. But it is step four, not step one.
The Founder Bottleneck explains why this sequence matters structurally. Outsourcing AI to an agency without extracting IP first does not remove the bottleneck. It adds a dependency to it.
The Capability Asset
The underappreciated value of done-with-you is what it builds beyond the AI system itself.
A founder who goes through a done-with-you program emerges with three things a done-for-you engagement does not produce:
Three capability assets
Documented IP
The methodology, fully extracted and structured. This has value independent of AI: for hiring, for training, for quality control, for eventually selling the business.
- Hiring. New team members can deliver your methodology from the documentation
- Quality control. Standards are explicit and measurable, not implicit and variable
- Business value. Documented IP increases the value of the business itself
These are assets that compound. The documented IP becomes more valuable as it is refined. The system literacy enables better decisions about every future AI investment. The iteration velocity means the business moves faster with each cycle.
20%
higher profitability in companies that centralise and structure their IP and data approaches
A done-for-you build delivers a system. A done-with-you program delivers the infrastructure for ongoing AI leverage. For most $500k+ knowledge businesses, the infrastructure is worth more than any single system.
If you want to assess which model fits your situation and what needs to be true before either one delivers lasting value, the IP Monetisation Assessment maps your current IP readiness and identifies the right starting point for your business.
For the full framework on AI transformation readiness, read AI Transformation Readiness. It covers the preconditions that determine whether any AI engagement succeeds or fails.