Maps client workflows and keeps every build accountable to an operational outcome, not a demo.
ZeptazAbout Zeptaz
The people behind the system.
AI automation and phone agents for service businesses. Operator-first, human-approved, and built to be reliable inside your stack.
01 / Our story
Reliable systems over hype.
We started Zeptaz because the same thing kept breaking inside good service businesses: work entered from everywhere, then got stuck between tools. Inquiries lost an owner, records fell behind, and follow-ups slipped.
So we build the boring, load-bearing systems that keep work moving. Connected workflows across intake, sales, onboarding, marketing operations, and reporting, with AI kept inside a controlled, human-approved loop.
We are operator-first and specific. We work in workflow maps, examples, and operational consequences, not jargon or guaranteed-revenue claims.
Fig. / The pattern we kept seeing
02 / Milestone
Our product won IgnitX 2026. HUTCH Resolve, chosen from 48 teams.
HUTCH challenged undergraduate teams to turn real customer problems into practical digital products using AI and automation. We entered as Zeptaz and built HUTCH Resolve - an evidence-backed complaint-resolution engine - on the same rules we use for every client system.
01 / TEAMS ENTERED
48
02 / FINALISTS
5
03 / CHAMPION
1
ZEPTAZ
- ORGANISER
- HUTCH
- TRACK
- Track A · Resolve & Support
- FINALS
- 5 Oct 2026
- RESULT
- Winner · 1 of 48
HOW IT WENT
- 11-19 SEPRegistrationHUTCH opens IgnitX to undergraduate teams across four challenge tracks.
- 21 SEPOrientationWe choose Track A: Resolve & Support - customer complaints about balance, data, packages and charges.
- 28/29 SEPIdea pitch48 teams pitch. Five go through to the build phase - Zeptaz is one of them.
- TO 4 OCTBuildA working product: trilingual chat, a voice channel, the investigation engine and an agent dashboard. v1.0.1 tagged on 4 Oct.
- 5 OCTFinalsThe five finalists present. HUTCH Resolve takes first place.
WHAT IT PROVES
- 01
The same rules as client work
AI for language, code for every number and decision, a human when the evidence runs out - the principles on this page, tested under competition pressure.
- 02
Our voice stack, in a new market
The voice channel runs on Zeptaz Voice - the same engine behind our AI phone agents - in English, Sinhala and Tamil.
- 03
Two weeks, end to end
From orientation to a live, tested demo at resolve.zeptaz.com, documented and versioned like every system we ship.
03 / Operating principles
Six things we refuse to be.
Every build gets held against these. They cost us some projects - that’s the point.
01
Not lead generationWe fix the work you already have
We don’t promise new leads or cold outreach. We improve how inbound inquiries get handled, owned, and followed up.
02
Not random zapsA connected system, not a menu
Modules depend on each other. You size the build - you don’t cherry-pick load-bearing steps that break the system.
03
Not autonomous AIAI stays human-approved
AI drafts and sensitive outputs are prepared for human review. No autonomous sending, no candidate ranking, no hiring decisions.
04
Not broad accessMinimum access, data-safe
Least-privilege by default: no raw password sharing, anonymized or test data where possible, clear PII handling.
05
Not hype-firstOperator-first and specific
We work in workflow maps, examples, and operational consequences - not jargon or guaranteed-revenue claims.
06
Not duct tapeReliable systems over one-offs
A maintained workflow engine beats a drawer of brittle one-off automations that quietly break.
04 / The team
People behind the system.
A small, hands-on founding team in Sri Lanka. We build the first systems ourselves - and document every reusable module as we go.
Designs the connector layer - the unglamorous plumbing that keeps records in sync.
Owns testing and reliability - the pass/fail scenarios and punch list nothing ships without.
Builds the engine’s automation logic and the human-approval loops around the AI.
05 / FAQ
Common questions.
Straight answers on scope, data, and how the work runs. Still unsure? A fit call sorts it fast.
Intake, routing, CRM/ATS updates, follow-ups, handoffs, approvals, reporting, and documents across your service workflows.
No. The system reduces admin chasing and prepares drafts, tasks, alerts, and summaries. Your team stays in control.
Not by default. AI drafts and sensitive workflow outputs are prepared for human review and approval.
Only if access and integration are practical. Projects must pass compatibility and access checks first - connector-first, then API where needed.
Minimum required access, no raw password sharing, and anonymized or test data where possible.
Get started
Work with us.
Book a workflow fit call and we will map where work is leaking, then recommend a build scope, timeline, and price range.