The short answer
Choose DIY when your team has the time, skill and appetite to build and maintain each agent. Choose managed when you want the conversation layer delivered and improved without adding delivery headcount. Many agencies end up with a mix: DIY for simple clients, managed for complex or high-value ones. Neither model removes the need for client-approved information and a clear human handoff.
Your agency decides to add AI conversations to its client offer. The demo looks easy: paste in a website, get a working agent. Then the real questions arrive. Who checks the answers for accuracy? Who updates it when the client changes pricing? Who reads the conversations that go wrong?
Those questions, not the feature list, usually separate a good deployment from a frustrating one. This guide compares the two delivery models fairly so you can match each to the client in front of you.
Publisher disclosure: This guide is published by ConvoRev, which offers managed implementation. The comparison is a decision framework, not an independent benchmark, and DIY is the right choice in many situations.
Related reading: choosing client communication automation and how to automate lead response for clients.
What each model actually means
DIY automation
You choose a platform and your team does the work: configure the agent, gather knowledge, write qualification flows, connect channels, test, monitor and update. The provider supplies software and documentation, and sometimes support, but delivery sits with you.
Managed AI agents
A provider handles discovery, knowledge preparation, agent design, channel and workflow setup, testing, launch monitoring and ongoing optimization under an agreed scope. You and the client still approve the content and own the relationship.
There is a middle ground too: platform features you configure yourself with onboarding help, or a consultant who builds the first version and hands it over. What matters is knowing exactly which tasks are yours.
The work that exists either way
Whatever you choose, a reliable client deployment includes the same jobs. The only question is who does each one.
- Discovery: understand traffic sources, common questions, the conversion event and current tools.
- Knowledge preparation: collect, clean and resolve conflicts in the client’s information, then get it approved.
- Conversation design: voice, qualification questions, boundaries and escalation rules.
- Integration and handoff: connect calendars, CRMs or notifications and test failure cases.
- Testing: normal, ambiguous, outdated, off-topic and escalation scenarios.
- Launch monitoring: read early conversations and fix problems quickly.
- Ongoing review: keep answers current as prices, offers and policies change, and improve from real conversations.
- Reporting: connect activity to qualified opportunities and outcomes where attribution allows.
An auto-generated first draft, such as one built from a client’s website, speeds up the early steps. It is a starting point, not a finished deployment. Testing, escalation design and review still need a person.
Side-by-side comparison
| Factor | DIY automation | Managed AI agents |
|---|---|---|
| Who does the work | Your team builds, tests and maintains each agent. | The provider handles the delivery tasks in the agreed scope. |
| Up-front cost | Often lower platform cost, with staff time as the main expense. | Usually includes an implementation fee for setup and testing. |
| Ongoing cost | Platform and usage fees plus the labor of reviews and updates. | Recurring management fee plus usage and channel charges, scoped by agreement. |
| Speed to first agent | Fast for a simple draft; slower to a well-tested one. | Depends on discovery and client approvals, but with a defined process. |
| Control and customization | Maximum direct control over every setting and flow. | Control through scoping and approvals; changes go through the provider. |
| Quality over time | Depends on whether anyone has time to review conversations. | Review and optimization are part of the service. |
| Scaling across clients | Each client adds setup and review hours to your team. | Capacity depends on the provider’s process and your commercial terms. |
| Accountability | You own results and errors end to end. | Shared: the provider owns delivery; you own the client relationship and approvals. |
Costs and inclusions vary by provider. Confirm current pricing and scope before you quote a client.
When DIY is the better fit
- You have staff who enjoy conversation design and can reserve time for testing and weekly review.
- Clients have simple, stable offers and low risk from an imperfect answer.
- You want deep control over configuration and are comfortable owning every error.
- You are experimenting with an offer and want to learn the mechanics before committing to a managed model.
When managed delivery is the better fit
- You want to add the service without hiring or reassigning delivery staff.
- The client has complex products, multiple qualification paths or sensitive topics that need careful escalation design.
- Inquiry volume is high enough that an inaccurate or stalled agent has a real cost.
- You want regular conversation review and knowledge updates as part of the service rather than a task nobody owns.
- You need a partner who has already worked through the testing and handoff details for similar businesses.
The costs people forget to count
A low software price can hide a lot of labor. When you compare options, account for these:
- Setup hours for discovery, knowledge cleanup and approvals, per client.
- Testing time for hard scenarios and failed handoffs, not just the happy path.
- Review time each week to read conversations and correct errors.
- Update time whenever pricing, hours, offers or staff change.
- Usage and channel charges such as AI processing, messaging and phone numbers, which vary.
- Integration work to connect the CRM, calendar or notifications and maintain the connection.
- Risk time for fixing a wrong answer before it damages the client’s trust.
Put a realistic hourly value on the labor you will provide. Sometimes DIY still wins. Sometimes a managed fee is lower than the staff time it replaces. The comparison only works if both sides are counted.
Seven questions to decide for each client
- Who will read the conversations every week, and do they actually have the time?
- How costly is a wrong answer for this client’s business?
- How complex is the qualification and handoff? One next step, or several paths?
- How often does the client’s information change, and who will tell us?
- What integrations are needed, and who maintains them?
- What are we promising the client, and can we deliver it with the resources we have?
- If the person who built this leaves, can someone else take over?
A client with a simple offer and low stakes may suit DIY. A client with a complex offer, higher inquiry value and little internal time may suit managed delivery. Treat it as a per-client decision rather than an agency-wide rule.
What to ask a managed provider
If you consider managed delivery, ask for specifics before you sign:
- Exactly what is included in setup, and what counts as a scope change?
- How often are conversations reviewed, and what do you receive afterward?
- How are knowledge updates requested and how quickly are they made?
- Which integrations are tested and supported, and which need custom work?
- What usage and channel charges apply, and how are they billed?
- How is client data kept separate, and who can access each deployment?
- What happens to the content and configuration if you end the relationship?
- What does the provider refuse to promise about results?
Be wary of anyone who guarantees leads, appointments or revenue, who calls the service fully autonomous, or who says it works with every platform. Honest providers describe the integration method and the limits of the system.
Frequently asked questions
What is the difference between a managed AI agent and DIY automation?
With DIY, your team builds, tests, monitors and maintains the agent on a platform you configure. With a managed agent, a provider handles discovery, knowledge preparation, configuration, testing and ongoing optimization under an agreed scope.
Is DIY cheaper for an agency?
Not necessarily. Software may cost less up front, but setup, testing, monitoring and update time is a real cost. Compare total delivery cost across platform fees, usage, labor and review.
Can we start DIY and move to managed later?
Often yes, especially if knowledge, qualification rules and handoff logic are documented. Ask any provider about migration and who owns the content before you commit.
Does a managed service guarantee results?
No. A managed service should provide structured setup, testing and review, but no provider can responsibly guarantee leads, appointments or revenue. Those also depend on traffic, offers and the client’s follow-up.
Does managed mean the agency steps out of the picture?
No. You stay the client’s strategic partner. The client or your agency still approves knowledge and claims, and a person still owns escalated opportunities.
Can we use both models across different clients?
Yes. Many agencies choose per client based on complexity, stakes and available time rather than committing the whole book to one approach.
See the conversation for yourself
Not sure which model fits your next client?
Open the ConvoRev chat widget on this page and ask: “What would managed implementation cover for a client with after-hours inquiries?” Then tell us about the client you have in mind and we can talk through the right approach, including when DIY makes sense.
Talk through your agency’s options →Explore ConvoRev and current plans
If the widget does not load, use the email link above. Channel availability, plan features and integration requirements vary. Managed services may be scoped separately.
Sources and further reading
- ConvoRev — current plans and managed implementation information.
- AI lead response — capability overview.
- Lead qualification — capability overview.