FullSession | Fin AI alternatives: compare your helpdesk, workflows and billing model | Roman Mohren | October 2026
Article: https://www.fullsession.io/blog/fin-ai-alternatives

AI AGENT SHORTLIST WORKSHEET
A working tool for replacing or comparing an AI support agent. Every field is blank on purpose: fill it in from your own setup and from each vendor's own pages. Nothing here reports a completed test or a vendor result.

HOW TO USE THIS FILE
Part A decides your path. Part B records the setup you have today. Part C is the shortlist grid, one block per finalist. Part D tests the jobs. Part E maps your own conversations to each vendor's billing unit. Part F is the decision record.
Vendor prices, units and minimums change. Write the URL and the date you checked next to every figure, and write "not published" when a figure is missing.


PART A: WHICH PATH ARE YOU ON?

Does your current helpdesk work for your team (queues, routing, reporting, seats)? (yes / no / partly):
If partly, what is not working:
Path chosen (circle one):
  A. Keep the helpdesk and change the AI agent
  B. Buy a desk with the AI agent included
Reason for the choice, in one sentence:
Who approved the path:


PART B: YOUR SETUP TODAY

Helpdesk name and plan:
Helpdesk cost per month (seats x rate, or plan price):
Channels the AI agent must cover (web chat, in-app, email, messaging apps, voice, other):
Languages:
People who take over from the agent (count, roles):
Office hours and time zone:
Current AI agent and how it is billed:
Last month's billed AI units (for Fin: outcomes):
Last month's minimum or commitment, if any:
Sales or qualification use on the current agent (yes / no):
Onboarding or in-product guidance use (yes / no):


PART C: SHORTLIST GRID (copy this block once per finalist)

Agent:
Path it fits (A / B / both):
Runs in my helpdesk? Source URL and date checked:
Channels it covers in my helpdesk:
What my helpdesk plan must include for it to connect:
Where people work after a handoff:
Knowledge sources it answers from:
Account data and actions it can reach, and who approves an action:
Billable unit, in the vendor's own words:
When a charge starts:
Is a handoff billed? Under what condition:
Is follow-up in the same conversation billed again:
Published price per unit (or "not published"):
Monthly minimum or commitment:
Seat or agent fees:
Other fees named on the pricing page:
Source URLs and date checked:

For FullSession Max, the current terms are on https://www.fullsession.io/pricing?solution=desk and https://www.fullsession.io/solutions/support-sales-desk. Check them on the day you fill this in.


PART D: JOBS TEST (run in the helpdesk or desk you will actually use)

Mark each result: pass / partial / fail / not tested. Note the configuration it needed and the evidence you saw.

1. Support answer in the web app, from an approved source
   Agent:            Result:            Configuration needed:            Evidence:
2. Supported task completed inside the app (state: completed / failed / unknown)
   Agent:            Result:            Configuration needed:            Evidence:
3. Website buying question that ends in a plan choice or a demo request
   Agent:            Result:            Configuration needed:            Evidence:
4. New-user onboarding step, then a return visit (is progress remembered?)
   Agent:            Result:            Configuration needed:            Evidence:
5. Customer asks for a person (does the context arrive with the conversation?)
   Agent:            Result:            Configuration needed:            Evidence:
6. Question the agent should decline (restricted or missing information)
   Agent:            Result:            Configuration needed:            Evidence:
7. Typed sensitive input (is it masked or redacted as your policy requires?)
   Agent:            Result:            Configuration needed:            Evidence:

A privacy failure, or an action reported as complete when it was not, counts as a fail.


PART E: UNIT MAPPING ON YOUR OWN CONVERSATIONS

Pick 20 recent conversations. For each vendor, mark how its own billing definition would treat the conversation: B = billed, N = not billed, U = unclear. Ask the vendor to confirm every U in writing. Do not convert one vendor's unit into another's; classify each conversation separately under each definition.

#   Short description                     Vendor 1   Vendor 2   Vendor 3   Notes
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

Totals (B / N / U) per vendor:
Vendor confirmations received for U items (date, name):
Forecast per vendor = billed units x published rate, then apply the minimum (show the arithmetic):


PART F: DECISION RECORD

Date:
Path:
Finalists:
Jobs covered by the chosen agent (from Part D):
Deployment confirmed in our own helpdesk or desk (yes / no, evidence):
Forecast monthly AI charge and the assumptions behind it:
Helpdesk or seat cost that stays on the bill:
Open questions and who owns them:
Decision and owner:
Date to review the decision: