AI Chatbot for Software Agencies: Qualify, Support, Resell
A dev shop has three separate reasons to run chat. Enquiries arrive from time zones where nobody on your team is awake. Products you maintain generate support questions you did not budget for. And clients keep asking you to put chat in the thing you built them.
The short version
- Engagement type is the first question. Build, rescue, augmentation or maintenance.
- Never automate an estimate or a security questionnaire. Both bind you later.
- Removing the EzyConn badge comes with the Professional plan. Know whether the client wants it.
- One chatbot per product. 1 on Free, 3 on Starter, 10 on Professional.
Somebody is always awake, and it is not you
Software agencies sell across borders. A founder in another time zone reads your case studies at what is midnight for you, wants to know whether you take projects their size, and finds a contact form. By the time you reply, they have had three other conversations.
Chat does not fix the time difference. It removes the dead hours in between. The visitor gets your published answers immediately, leaves qualified details, and arrives in your inbox as a real enquiry rather than a name and a one-line message. Answers come back in many languages, which matters more for a distributed client base than most agencies expect.
Qualify on engagement type first
Agencies waste more time on badly sorted leads than on lost pitches. A rescue job and a greenfield build need different people in the room, and staff augmentation is a commercial conversation rather than a technical one. Sorting that in chat costs the visitor thirty seconds.
| The bot asks | What it tells you |
|---|---|
| New build, rescue, augmentation or maintenance? | Four conversations, four owners on your team |
| What is it built on today? | Rules out the stacks you will not take on |
| What budget range are you working to? | The fastest way to end a mismatch politely |
| Is there an in-house team already? | Decides whether you lead or plug in |
| When does this need to ship? | Separates a funded project from an idea |
Publish rate bands or a project minimum and let the bot repeat them. Agencies resist this, then spend two calls a week discovering the same mismatch by hand. A range filters politely and costs you nothing you were going to win.
What the bot must not touch
Two things need saying clearly to whoever configures this. The bot never estimates, and the bot never answers a security or compliance questionnaire. An estimate given in chat becomes an expectation, and a security answer given in chat becomes a claim you have to stand behind.
| Message | Handling | Why |
|---|---|---|
| How do you run discovery? | AI answers | Published process |
| Do you work in our stack? | AI answers | Published capability list |
| Who owns the code at the end? | AI answers | Your standard terms |
| Can you estimate this build? | Escalate | Needs an engineer, not a guess |
| Fill in our security questionnaire | Escalate | Never automate a compliance answer |
| Our production app is down | Escalate now | Straight to your on-call path |
The last row matters most. If you maintain production systems, an outage message arriving in a chat widget must reach a human immediately, not sit in a queue. Set that route up on day one and test it with a fake incident.
The support load nobody quotes for
Every agency that keeps clients after launch ends up doing support it never priced. How do I add a user. Why is this report empty. Where do I change that setting. These are documentation questions arriving as developer interruptions, and they are exactly what an AI chatbot handles well.
Point it at the product documentation you already wrote and let it answer the repeating ones. Real bugs and anything about data still go to a person. The same split works whether you are supporting a client product or your own, and we go deeper into it in our guide to AI chat for technical support. If the product is a SaaS with a signup flow, the onboarding side is covered in our SaaS onboarding guide.
Reselling: the limits, stated plainly
Branding is where agencies trip up. Removing the EzyConn badge, so a client product shows only the client brand, comes with the Professional plan. If you want the chat to read as your agency platform rather than a neutral widget, that is the tier.
| Plan | Price | Chatbots | Messages a month | Notes |
|---|---|---|---|---|
| Free | $0 forever | 1 chatbot | 500 | Trial on one site |
| Starter | $25/mo | 3 chatbots | 5,000 | Three client products, 3 agent seats plus the owner |
| Professional | $95/mo | 10 chatbots | 25,000 | Remove the EzyConn badge, 10 agent seats plus the owner, AI Insights |
One chatbot covers one client product, so a single account reaches ten clients at most. Past that, each client gets their own account, which is usually the better arrangement anyway. They own the bill and the history, and you charge for setup, content training and monthly tuning. Annual billing takes 20% off, and the details are on the pricing page. Clients with procurement teams should look at the Enterprise tier.
Where this earns its keep
Licence markup is not a business. The work clients cannot do is the valuable part: writing the answers, deciding what escalates, wiring handover into their tools, and reviewing transcripts each month. EzyConn trains by crawling the site or product docs, GPT-4o and Claude run the AI, and Zapier, Slack, Microsoft Teams and WhatsApp work on every plan, Free included. Add a chatbot to the next maintenance retainer, keep an hour a month for tuning, and you have a line item that renews.
Frequently asked questions
What should a software agency chatbot qualify for?
Engagement type first, because everything else follows from it. A new product, a rescue of an existing codebase, staff augmentation and ongoing maintenance are four different conversations with four different people on your team. After that, ask about the stack, the budget range, the timeline and whether an in-house team already exists. Five answers is usually enough to decide whether a call is worth booking.
Can the chatbot answer technical questions about our services?
It can answer what you publish: the languages and platforms you work in, how your discovery phase runs, how you handle code ownership and handover, and what your support terms look like. It should not design an architecture, estimate a build, or fill in a security questionnaire. Those need a named engineer, and a wrong answer in chat can follow you into a contract.
Can we white label the chatbot for client products?
Yes, on the Professional plan. Removing the EzyConn badge comes with Professional, so a client product carries the client brand in the chat rather than ours. Choose based on whether the chat is meant to look like the client product or like your agency platform.
How many client products can one account cover?
One chatbot per product, and the limits are fixed. Free gives one chatbot, Starter at $25 a month gives three, and Professional at $95 a month gives ten. Beyond ten products, put each client on their own account. That also keeps their conversation history with them if the engagement ends.
How do we size message limits for a client product?
Size against the busiest month of the busiest product, not an average. An agency marketing site might use a few dozen messages a month, while a client SaaS app with real users can use thousands. The caps are 500 messages a month on Free, 5,000 on Starter and 25,000 on Professional. Annual billing takes 20% off.
Start on your own site
The free plan is $0 forever, no credit card: 2 seats, 500 messages a month, one widget, and the mobile app. Run it on your own site before you sell it to anyone.
Start freeLast updated . EzyConn plan details, chatbot limits and white label availability are current as of August 2026. View more guides.