Should your AI agency sell custom services, productized offers, or both? Here's what I learned running Allwebzone and why the answer is not obvious.
When I started Allwebzone, I sold custom AI consulting. Every client got a bespoke solution. Every project was unique. Every proposal was written from scratch.
Revenue was good. Margins were not.
The problem with pure custom services is that you sell your time. Even with AI making you faster, every new client requires new discovery, new architecture, new implementation. You cannot scale your time.
Then I watched agencies in other industries (web design, marketing, accounting) solve this by productizing. Fixed-scope packages at fixed prices. Less customization, more repeatability. Better margins.
I tried going fully productized. That did not work either.
Here is what I landed on.
Custom AI consulting feels premium. Every solution is tailored. Clients feel special. You charge high rates because the work is unique.
The problems emerge at scale.
Scope creep is constant. When everything is custom, the boundaries between “what we agreed to” and “one more thing” blur. Clients expect flexibility because they are paying for a bespoke engagement.
Knowledge does not compound. The solution you build for Client A rarely transfers to Client B. Different industries, different workflows, different technical environments. You start each project roughly from zero.
Pricing is hard. Every project needs a unique estimate. You spend hours scoping before you know if the deal will close. Win rates on complex proposals hover around 30%, which means 70% of your scoping time is wasted.
Revenue is lumpy. Big projects come and go. One quarter you are flush. The next you are scrambling. Recurring revenue is minimal because custom projects end.
I ran Allwebzone in pure services mode for the first eight months. Revenue grew. But so did the operational overhead of managing unique engagements.
The exact architecture, memory layers, and delegation patterns I use to run 50 agents across two businesses.
Get the AI Agent Blueprint →The obvious fix: productize. Create fixed packages with defined scopes, deliverables, and timelines. Sell them repeatedly.
I tried this. Three tiers: workflow automation, intelligent systems, and full AI architecture. Clear pricing. Clear scope. Clear deliverables.
Two things went wrong.
Clients did not fit the boxes. The prospect who needed 80% of Tier 2 but also one feature from Tier 3 was common. The prospect who needed something between Tier 1 and Tier 2 was even more common. Rigid tiers lost deals because the fit was not close enough.
Premium clients went elsewhere. The best clients, the ones with complex needs and big budgets, did not want a package. They wanted a partner who understood their specific situation. A productized offer felt like being fit into a template, which is exactly what it was.
I was losing deals at both ends. Smaller clients wanted more flexibility in the packages. Bigger clients wanted more customization than packages allowed.
After a year of experimentation, here is the model that stuck.
Productized entry points. Three fixed-scope packages that solve common, well-defined problems. Email automation. Content pipeline setup. Client onboarding automation. Each package has a clear price, a defined timeline, and standardized deliverables.
These packages are not the goal. They are the door. They get clients in the door with a low-risk purchase. Scope is tight. Expectations are clear. Delivery is efficient because I have done the same thing many times.
Custom expansion. Once the productized engagement is complete and the client sees results, the conversation shifts naturally. “This email automation is great. Can you do something similar for our reporting?” That becomes a custom engagement, scoped with full knowledge of the client’s environment and needs.
The productized entry point earns trust and demonstrates capability. The custom expansion captures the higher-value opportunity.
Retainer for ongoing. Both productized and custom clients can add a monthly retainer for monitoring, maintenance, and iteration. This creates recurring revenue and long-term relationships.
The revenue mix is roughly 30% productized, 50% custom, 20% retainer. The productized work is the most efficient. The custom work is the most profitable per project. The retainer is the most predictable.
Building a good productized offer requires knowing which problems show up repeatedly. After fifty-plus client conversations, clear patterns emerged.
Most businesses waste time on email. Not “email marketing” but internal email management. Triage, follow-ups, scheduling. The email automation package practically sells itself because everyone feels this pain.
Content is the universal bottleneck. Every business knows they should produce more content. Few have the time or process to do it consistently. A content pipeline that produces three posts per week without manual effort is immediately valuable.
Onboarding is always messy. New clients or new employees, the onboarding process is usually a mix of manual steps, forgotten emails, and inconsistent experiences. Automating it creates a professional first impression that compounds through referrals.
Each package started as a custom project for a client. When I had done the same project three times with minor variations, I standardized it into a package with documentation, templates, and a delivery playbook.
Productized packages: $2,500 to $5,000. Fixed price, no negotiation. The scope is the scope. If the client needs more, that becomes a custom engagement.
Custom projects: $7,500 to $50,000+. Priced on value, not cost. The scoping conversation establishes the problem, the impact of solving it, and the investment required. Higher-budget clients get the full architectural treatment.
Retainers: $500 to $2,000/month. Based on complexity. A simple automation needs minimal monitoring. A multi-agent system needs active maintenance.
The key pricing insight: productized packages build your reputation. Custom work builds your revenue. Retainers build your stability. You need all three.
Running both productized and custom work requires different operational modes.
Productized work is systemized. There is a delivery playbook. The agents follow the same steps every time. Quality is consistent. Timelines are predictable. I can delegate most of the delivery to agents with minimal oversight.
Custom work requires more judgment. Each engagement has unique elements. The agents handle the repeatable parts (research, initial drafts, standard integrations), but the architecture decisions and client-specific adaptations are human work.
The agents make the productized side possible and the custom side more efficient. Without them, the productized packages would eat too much time to be profitable at their price point. And the custom work would be too slow to deliver competitive timelines.
Do not start with products. You need to do custom work first to understand which problems repeat. The patterns that become packages only emerge after ten to fifteen client engagements. Trying to productize on day one means guessing, and you will guess wrong.
Do not stay in custom forever. At some point, the repeat patterns become obvious. When you are scoping a project and thinking “I have done this exact thing four times,” that is your signal to productize.
Build the internal tooling early. The agents and systems you build for yourself are what make the hybrid model work. If you are manually delivering productized packages, the margins disappear. If you are doing custom work without any agent assistance, the timelines slip.
Measure what matters. For productized packages: delivery time, client satisfaction, and gross margin per package. For custom work: scope accuracy, profit per project, and conversion from productized to custom. For retainers: churn rate and average lifetime.
The newsletter covers the business side of running an AI agency, including real numbers and operational lessons.
Services or products is a false choice. The answer, for most AI agencies, is a deliberate combination that matches how clients actually buy.
Subscribe to The Signal
One issue per week. The biggest AI build of the week, the top five videos worth your time, and an operator's read on what actually matters. For builders who ship, not dreamers who scroll.
No spam. Unsubscribe anytime. Just The Signal.