A short tour through the naming decisions for our AI surface — what we kept, what we threw out, and the three rules every product name has to pass.
A product name is the cheapest piece of the product to change in theory and the most expensive in practice. Once a name is shipped to customers — in dashboards, in invoices, in documentation, in support transcripts — it gets stitched into the product's identity in ways that show up two years later. Renaming "Workspaces" to "Projects" is the kind of thing that costs a quarter of internal communications work, leaves stragglers in the codebase for years, and confuses customers who have a six-page runbook that references the old name.
So we put real time into names. Not days; weeks. And we have three rules that any candidate name has to pass before it ships.
If a sales engineer has to start a demo with "this feature is called X, but what it actually does is Y," the name has failed. The name has to do the heavy lifting on the dashboard, in the docs, and in the marketing page before any human gets to explain it.
Names that work as labels often fail as nouns inside sentences. "Configure your IntelliFlow" is a sentence that a customer hears as a brand-jingle phrase, not as an instruction. "Configure your dial plan" is a sentence that sounds like normal English. We test every name in three sentences before shipping it: "How do I [verb] my [name]?" "Where do I find [name]?" "[Name] is broken." If any of the three sounds embarrassing, the name is wrong.
A name implies a scope. "AI Coach" implies the AI is coaching the agent. If, in practice, the AI is summarising calls for the supervisor and never speaks to the agent, "AI Coach" is lying. Lying names erode trust slowly. Every customer who clicks the AI Coach tab expecting a coach and finds a summariser will, for the rest of their time as a customer, distrust the next name they see.
The headline naming decision of the past 18 months was what to call our flagship AI surface. The product handles inbound calls — it answers, routes, books, takes payment when appropriate, transfers when necessary. The candidate names on the whiteboard at one point exceeded 60. We narrowed to four serious finalists:
Receptionist was the front-runner for most of the process. It satisfied rule 1 effortlessly. It satisfied rule 2 in most sentences. It failed rule 3 in a way that took us months to admit.
"Receptionist" implies a specific person at a specific desk doing a specific job — taking calls and routing them. Our product does more than that. It books appointments, qualifies leads, takes payment, sends follow-up texts, transfers to humans when needed, and handles outbound reminders. A receptionist does not do most of those things. Calling the product a receptionist would have invited a customer expectation that the product was narrower than it actually was. The lie was small, but it was a lie.
We tested "Voice" against the rule set instead. Rule 1: a customer hearing "AI Voice" for the first time imagines, roughly correctly, "this is a thing that handles voice interactions for my business." It is not perfectly descriptive, but it is not misleading either. Rule 2: "configure your Voice" reads as natural. "Voice is busy" reads as natural. "Voice missed a call" reads as natural. Rule 3: Voice promises voice handling. The product handles voice. The name does not over-claim.
Voice won. The runner-up, Receptionist, lost on a single rule and was otherwise excellent.
A short list of candidate names that did not survive the rules, with the lesson each one carried.
SmartAnswer. Killed by rule 2. "Configure your SmartAnswer" reads like a brand jingle. Words ending in -er are awkward as product names; words concatenated with "Smart" are worse.
HelpDesk AI. Killed by rule 3. Our product is not a help desk. A help desk is a ticketing system. Calling the AI a help desk would have invited every customer to expect ticket management.
AjoxiBot. Killed by rule 1 in a particular way: customers who knew us would understand it, but customers who did not would hear "bot" and downgrade their expectations. The product is more sophisticated than "bot" suggests, and the name was actively undermining the demo.
Caller. Killed for being unsearchable. A name that is also a common English word in your product's domain is a documentation nightmare. "How do I configure Caller for after-hours?" returns infinite false positives in any internal search. We added "searchability" as an informal fourth rule after this one.
We do not have a formal naming committee. We have a loose rule: any product name that will be customer-visible has to be reviewed by three people from outside the engineering team that built the feature — typically one from sales, one from support, and one from marketing. The reviewers are given the candidate name, two sentences of context about what the feature does, and asked to write down their first guess about what the feature does based on the name alone.
When the three guesses match the feature, the name passes. When they diverge, the name fails. The process takes 20 minutes and is the highest-leverage 20 minutes in any product launch.
We have run this on every customer-facing name shipped in the last four product cycles. About one name in four fails the first round and gets reworked. About one in eight fails twice. We have never regretted spending the extra week.
Naming is the part of product work that looks like procrastination from outside the room. "We are going to spend three weeks deciding what to call this." It looks self-indulgent. It often is.
But the name is one of the few decisions on a product that everyone who interacts with the product encounters, every time, forever. Getting it right is unspectacular and easy to underrate. Getting it wrong is slow, expensive, and the kind of thing engineers complain about in interviews about why they left their last company.
So we use the three rules. They are not magic. They are just three filters that catch the most common mistakes before the mistakes get printed on invoices.
AI receptionists, wholesale routes, virtual numbers — built on one platform with transparent pricing and a 24/7 NOC.