Every reply your agent writes is generated by an AI model, and you choose how that model is picked on the agent's Agent screen under AI Model. For most agents the best choice is one of the two Auto settings: eChat picks the model for you and keeps it current, so there's nothing to change on the agent as better models arrive.
This article explains the two Auto settings, how they choose a model and bill for it, when you might pick a specific model instead, and how to test a switch with evidence rather than guesswork.
Auto (Smart) and Auto (Efficient)
What it is
The AI Model picker lists two Auto settings first, under Recommended:
- Auto (Smart) - answers with the strongest model eChat has measured on real conversations. It's the default for new agents and the right choice for most agents.
- Auto (Efficient) - answers with the best lowest-cost model eChat has measured, for busy agents where the cost per response matters more than the last bit of quality. No option in the picker costs less per response.
The credits each one costs per response are shown next to it in the AI Model card. Auto (Smart) usually costs more than Auto (Efficient) — the card shows a range such as 1–3 credits, usually 2–3, because short greetings and thanks can cost less (see below).
How it works
- eChat picks the model. Each Auto setting runs on a model eChat has chosen for that price level. You don't see or manage which one; the picker shows the setting, not the model underneath.
- It moves only to a proven model. eChat only switches an Auto setting to a new model after that model has been measured against the current one on real webinar conversations and is at least as good. When it switches, agents on that setting follow automatically.
- You pay for the model that answered. Each response is billed at the credits of the model that actually wrote it, within the range shown for that Auto setting in the card.
- Short greetings and thanks cost less on Auto (Smart). When a visitor sends a message that asks for nothing — thanks, hello, goodbye, or a brief acknowledgement — Auto (Smart) answers it the same way Auto (Efficient) would, and bills 1 credit for that turn. Ordinary questions still use Smart's usual cost.
- An Auto setting can have a backup. If its model refuses to answer a turn, another measured model at the same price level answers instead.
The cost applies to every turn where the model runs - including a notify-moderators interaction card where the agent looks and decides to stay silent. Your monthly allowance is shared across all your agents, so a pricier setting on one agent leaves fewer credits for the others. See Plans, Credits & Usage.
Agents that were set to a model eChat has since retired were moved to Auto (Smart) automatically. If one of your agents now shows a different setting than you picked, that's why; you can change it in the AI Model card at any time.
Best practices
Start with the default, Auto (Smart). It gives you the strongest measured answers without having to follow model releases yourself, and a baseline to compare against before you change anything.
Choosing between Smart and Efficient
What it's best for
- Auto (Smart) - most agents, and any agent where answer quality matters most, such as sales conversations or complex product questions
- Auto (Efficient) - high-volume agents answering routine, well-documented questions, where a lower cost per response matters more than nuance
How it works
Changing the setting takes effect for the agent's next replies. It doesn't change what the agent knows - training, Instructions, and Grounding strictness all stay the same - only which model writes the answer. If answers are wrong because training is missing or outdated, a stronger model won't fix that; add or correct Q&A instead.
Best practices
- Estimate the cost before you switch: multiply your typical monthly reply count by the credits per response shown in the AI Model card, and compare it with your plan's allowance.
- Don't upgrade to paper over thin training - fix the training first, then decide whether a stronger setting still helps.
- Revisit the choice after big changes in volume; what made sense for a quiet agent may be expensive once traffic grows.
Picking a specific model
What it is
Below the Auto settings, the picker also lists specific models by name, each with its credit cost per response. The list changes as models are added and retired, so check the picker for the current options.
How it works
A specific model stays fixed: your agent keeps answering with it until you change the setting, and it doesn't move to newer models the way Auto does. Each response costs the credits shown for that model.
Best practices
Pick a specific model only when you have a clear reason, such as an Evaluate result showing it works better for your agent. Otherwise, Auto keeps you on a measured model without any upkeep.
Testing a switch before you commit
What it is
Before moving a live agent to a different setting, you can measure the difference instead of guessing.
How it works
Try a handful of real questions in Playground first to get a feel for tone. Then run Evaluate: by default a run answers with the agent's own model, and you can pick another setting or model for the run to compare, without changing the agent. Run the same set on both and compare the scores. Evaluation runs use credits too, based on what the run actually cost, so a comparison on a pricier setting costs more.
Best practices
Decide ahead of time how much of a score change you'd accept for a given saving. A small dip for a big drop in credits can be a good trade - but make it a deliberate choice, and check Chat Logs in the days after a switch to confirm real conversations still look right.