Legacy SaaS dossier · zendesk

Zendesk

CategoryCustomer support / ticketingZendesk Suite + Advanced AISite ↗

A shared inbox with macros and SLAs, sold as a Suite with an AI tier that bills extra for the LLM your agents already use in another tab.

01 · Data

Data the SaaS sits on

Inbound emailYours

Knowledge baseYours

Customers tableYours

StripeAPI

Internal Build Score

0.71

Feasibility of a useful internal substitute for the actually-used 20% of this SaaS, built with Claude and your own data layer.

IBS = 0.25·U + 0.25·C + 0.20·D + 0.15·I + 0.15·Mthis record · 71%

U
User adoption gap
weight 0.250.70
How much licensed surface area sits idle.
C
Core simplicity
weight 0.250.80
How small the actually-used 20% is when re-implemented honestly.
D
Data portability
weight 0.200.62
How clean the exit is, exports, formats, audit lock.
I
Integration burden (inverse)
weight 0.150.60
How few external systems must keep working through the SaaS.
M
Moat erosion
weight 0.150.78
How much the original differentiator has been commoditised by LLMs + cheap infra.

Editorial heuristic. Not investment advice, not a procurement recommendation.

03 · Feature usage

Commonly used vs. rarely touched

Commonly used

Email + chat → ticket queue
Assignee + status + priority
Saved replies / macros
SLA timers per channel
Searchable history + customer profile

Rarely touched

AI add-on tier billing per resolved ticket on top of seats
Sunshine CRM platform pretending to be a CDP
Explore, analytics built on a custom query language
Talk (voice), Sell (CRM), Guide (KB), every adjacent vertical as a SKU
Workforce Engagement, schedules + QA bolted on after acquisition
Per-light-agent / collaborator pricing oddities
Connectors that exist only on the highest tier

Build it yourself→

Internal build

Build it yourself

01 · Stack & flow

Postgres

Next.js

Claude API

Postmark

Twilio (voice / SMS)

Internal build map

Data in

Inbound email

Customer row

Agent layer

Reply drafter

Tagger

Logic

queue + assign

retrieve KB

draft + cite

SLA timer

escalate triggers

Outputs

Draft reply

Ticket tags

SLA alert

Build time: 7 weeks
Run cost: $6/seat·mo
vs. SaaS: $240K / yr

Build outline

Inbound email → tickets via Postmark MX
Agent console: queue, claim, reply, tag
Claude drafts first reply + categorises
SLA timer service + alerts to Slack
Knowledge base = same markdown repo as the wiki

02 · Claude prompt

First, map what you actually use (optional)

Step 1 · usage audit prompt

// Zendesk usage audit, scope the rebuild Input: 90 days of ticket volume by channel, the macro list with use counts, the trigger / automation list, and the SKUs on the contract. ## Do 1. Keep channels with real volume; drop any under 1% of tickets. 2. Keep macros used more than 5 times in 90d; the rest are dead. 3. Separate triggers with external side effects (email, webhook) from field-only triggers. ## Output JSON: { keep_channels[], keep_macros[], keep_triggers[], drop[] }.

Step 2 · rebuild the core

// Support reply drafter, internal agent console You see (a) one customer ticket (subject + body + history with this customer), (b) the agent's last 20 most-used macros, and (c) the relevant KB articles retrieved by the previous step. ## What you must do 1. Draft a reply that resolves the actual question, no boilerplate openers, no "Thanks for reaching out". 2. Match the closest macro by name (return it in `macro_match`). 3. Flag refund / cancel / legal triggers in `escalation_flag`. 4. Cite each fact to a KB article id in `citations[]`. ## What you do not do - Do not promise refunds, discounts, or policy changes. - Do not invent KB content; if none retrieved, say so in the draft. - Do not write closing pleasantries, the agent's signature handles that. ## Output JSON: { draft, macro_match, escalation_flag, citations }.

03 · Result

Charged twice for May. Please refund.

ticket:38192

Refund initiated on duplicate May charge ($29.00), Stripe ref ch_3Ru… · macro: billing-double-charge · escalation: false

← Legacy SaaS index AI wrappers index →