The report is due. The number is a guess.
Retention gets estimated from whoever picked up the phone, not from the full cohort.
DozAI Mission 01 · Pilot design cohort open
Know where every graduate is, 90 days later.
Your funder wants 3, 6 and 12-month retention numbers. Your trainers are already running the next batch. DozAI follows up with every graduate on WhatsApp and voice, in their language, verifies what has changed, and hands your team only the cases that need a person.
Your outcome definitions, not oursEvidence on every casePeople make the final call
184
Case 0184 · Retail associate cohortCompleted Mar 2026 · Placed: Bright Retail
● IN REVIEW
- Day 30 · WhatsAppSame jobStill with the placed employerYes, still at Bright Retail. Floor staff, morning shift.
- Day 90 · Voice call · 4 minNeeds reviewChanged job. Employer differs from your record.“I moved to City Mart in June. Better salary, same work.”
DozAI proposes Changed job, retained in sector. Your reviewer confirms.
- Day 180ScheduledNext check-in booked automatically
One case out of a cohort of hundreds. Illustrative example.
Sound familiar?
Placement is the easy number. Retention is where it falls apart.
You can tell a funder how many people you trained and placed. Ask what happened six months later and it becomes a scramble.
Half the phone numbers have changed.
Graduates switch SIMs, cities and jobs. One missed call and they drop off the list.
Calls happen between batches.
Trainers and placement officers squeeze in follow-ups, ask different questions and log answers differently.
Nobody can show the evidence.
When a funder or auditor asks “how do you know?”, the answer lives in a spreadsheet and someone’s memory.
The people who most need a follow-up, the ones who lost the job, are the ones you’re least likely to reach.
How the mission runs
You set the rules once. DozAI runs every checkpoint.
The mission works through your cohort at 30, 90 and 180 days (or whatever schedule your funder uses), and keeps going until every case reaches a clear state.
Load the cohort
From your programme tracking system or a spreadsheet. Consent and contact rules travel with each record.
Reach out
WhatsApp first, then a voice call in the graduate’s language. Retries within the contact windows you allow.
Ask what matters
Your questions, with follow-ups that branch: employer, role, start date, income band, reasons for leaving.
Check against records
Spots when answers contradict the placement on file, or when something sounds off.
Propose an outcome
Classifies each case and attaches the evidence. Clear cases are ready to approve.
Your team decides
Reviewers see only conflicts, sensitive answers and unclear cases. Approved results are saved back to your system.
DozAI does the workA person makes the call
What you get back
Every graduate lands in one clear state.
No more “couldn’t reach” as a silent default. Unreachable is a result too, with the attempts on record.
Same jobStill with the employer you placed them with
Changed jobNew employer, role or way of working
Not employedLooking for work, unavailable, or another reason, in their words
Education or trainingFurther study, apprenticeship or another course
Unreachable or unclearEvery attempt logged, or flagged for a person to look at
A case packet, not a call log
Each graduate’s result comes with everything a funder or auditor would ask for.
- Call recording and transcript, or the WhatsApp thread
- Structured answers: employer, role, start date, income band
- What changed since the last checkpoint, and why
- Who reviewed it, what they decided, and when
- A cohort view: retention by checkpoint, centre, trade and employer
And one thing most follow-up efforts never capture: the graduates who need help again. Anyone looking for work can be routed back to your placement team.
What your reviewer sees
Only the cases that need a person. With the evidence side by side.
On your record
Placed employerMetro Logistics
RoleWarehouse associate
CheckpointDay 90
What DozAI found
WHATSAPP · DAY 88“The company name you have is wrong.”
VOICE CALL · 00:42“I joined Metro Express in July. Same warehouse, they changed the contractor. I’m still working there.”
Proposed decision
RECORD CONFLICT
Graduate appears to be employed in the same role. The employer on file may need correcting.
ApproveEditAsk for evidenceReject
Every override needs a reason. Those reasons are how DozAI learns your organisation’s judgement, and they become test cases before any change launches. Names and details here are illustrative.
Tools, automation and human judgement
AI handles the volume. People handle the consequence.
We’re explicit about who does what. Nothing that affects a graduate’s record or your reported outcomes is decided without a person able to see why.
Repeatable, reversible work
- Scheduling checkpoints and retries
- Sending reminders and switching channels
- Running the interview in the graduate’s language
- Transcribing and pulling out the answers
- Honouring opt-outs immediately
Drafts a person can check
- Proposing an outcome for each case
- Flagging contradictions with your records
- Summarising why a graduate left a job
- Drafting the cohort retention report
Decisions with consequence
- Whether conflicting evidence counts as an outcome
- What meets your funder’s definition of “placed”
- Anything involving distress, safety or suspected fraud
- Changing the definitions themselves
WhatsAppFirst contact, reminders, quick replies
Zolabs voiceMultilingual AI calls for the last mile
Your systemZoho, CRM, MIS or spreadsheet, read and written back
Review workspaceEvidence, decision and reason in one screen
Audit trailEvery action, version and decision on record
Stage 1Shadow
Runs on past cases. We compare its answers with what your team decided.
Stage 2Draft
Works live, but a person approves every classification and update.
Stage 3Limited action
Sends reminders and clears simple cases. Unusual cases and a random sample still go to review.
Stage 4Managed autonomy
Handles defined case types end to end. Calls with real consequences stay with people.
DozAI starts as an apprentice. It earns autonomy task by task, only after it passes your review.
The pilot design cohort
We’re building this with a small group of skilling organisations.
Not a demo, and not a finished product. A real follow-up of one of your cohorts, with your team shaping the questions, the outcome definitions and the review screen alongside ours.
You run skilling or livelihoods programmesWith placement or self-employment as a stated outcome
You have a recent cohortGraduates who finished roughly 1 to 12 months ago
You can contact themPhone numbers on file and a consent basis for follow-up
You have a named ownerSomeone who can review cases a few hours a week
What you get
- The mission set up around your outcome definitions and your funders’ reporting requirements
- A supervised follow-up of a real cohort, on WhatsApp and voice
- Case packets with evidence for every graduate reached
- A retention report you can take to your board and funders
- A direct say in what DozAI’s first mission becomes
What we ask
- A cohort list and the placement records you already hold
- Your current follow-up questions and how you define “placed” and “retained”
- A reviewer’s time during the supervised weeks
- Honest feedback, including when something doesn’t work
- Your baseline: how long follow-up takes today, and what it costs
Week 1Mission contract
Agree definitions, questions, contact rules and what DozAI must never do.
Week 2Shadow
Test on past cases your team has already followed up. Fix the gaps.
Weeks 3–4Supervised live run
DozAI contacts your cohort. Your reviewer approves every outcome.
Weeks 5–6Limited autonomy
Clear cases flow through. We compare against your baseline and report.
What DozAI is
One mission. Built on a platform for many.
DozAI isn’t a placement-tracking app. It’s a platform for running missions: pieces of work with a clear end result, where AI does the repeatable work and people keep the decisions that carry consequence.
Placement and retention is the first. Every mission after it reuses the same foundation for cases, evidence, follow-up, review and audit, so your second mission doesn’t need a second system.
DozAI platformCases, evidence, scheduling, review and audit, shared across every mission
AI employeeA reusable role, like a verification associate, that collects and cross-checks evidence
MissionYour questions, contact rules, outcome rubric, messages and voice scripts
RunOne programme cohort, with the mission’s settings fixed so results are comparable
CaseOne graduate’s outcome, with every piece of evidence behind it
What comes next
How we choose a mission, and what’s on the list.
AI agents do well on work with bounded judgement: a known goal, a finite set of evidence, clear rules, and an obvious point where they must stop and ask a person. We test every candidate against the same criteria.
High volume · hundreds of similar cases
Repeats · every cycle or cohort
Evidence-based · records, documents, calls
A person decides · AI prepares, doesn’t rule
Named exceptions · most cases follow a known path
Reversible · errors can be corrected
Someone owns it · a team already pays for this work
Enrolment and attendance follow-up
Three missed sessions triggers a call. DozAI records the reason in the participant’s words and routes them to a staff member.
Safeguarding or disciplinary actionReferral closure and benefit receipt
Follows a health, education or welfare referral, or a distribution, until the service was received, refused or needs attention.
Closing a case when wellbeing is uncertainMissing documents and applications
Checks scholarship or programme applications for what’s missing, expired or inconsistent, and chases it. Before anyone scores merit.
Waivers and final eligibilityGrantee milestone evidence
Reads the grant agreement, builds the evidence checklist each reporting cycle, chases gaps and links every claim to a source.
Accepting weak evidence, releasing fundsClearing data errors
Runs data checks, lists duplicates and contradictions, fixes the obvious ones and proposes corrections for the rest.
Merging people’s records, changing outcomesSurvey back-checks
Spots impossible durations and identical answer patterns, picks a transparent sample for back-check calls and reports discrepancies.
Any finding about an enumeratorWhat we won’t automate
Award decisions for grants, scholarships or benefits. Safeguarding triage. Hiring and rejection. The volume is tempting, but the consequence for a person is too high for an AI to own.
Questions we’re asked
Before you apply
Does this replace our placement team?
No. It takes the repetitive calling off their plate so they can spend time on the graduates who need help: the ones who lost a job, or are struggling in one. Your team still owns every decision that goes into a report.
What about consent and our graduates’ data?
We agree the consent basis, what data the mission can see, and how long recordings are kept before anything runs. Graduates can opt out at any point and that’s honoured immediately. Your data stays within your organisation’s own secure space, and a lesson learned from your cases never moves to another organisation’s mission. Once the mission is complete, we can delete your data on request, including every call recording, transcript and message. We design this in line with India’s DPDP Act.
Which languages does it work in?
Voice calls run on Zolabs, EdZola’s multilingual voice platform, built for conversations across Indian languages. We’ll confirm the languages your cohort needs during the mission contract week.
What if a graduate doesn’t pick up?
DozAI retries within the windows you set, switches between WhatsApp and voice, and stops at the limit you agree. “Unreachable” is a recorded outcome with every attempt logged, not a blank row.
Which systems does it work with?
For the pilot, a spreadsheet export is enough. If you run on Zoho Creator or another system, we can read from and update it directly.
What does the pilot cost?
We’ll talk about it on our first call. The cohort is small on purpose and we want the right partners in it, not the ones who can pay most.
Pilot design cohort · Mission 01
Bring us one cohort. We’ll tell you where they are.
Tell us about your programme, roughly how many graduates you’d want to follow up, and when your next funder report is due. We’ll come back within a week.
Apply for the pilot cohort →
Please don’t send graduate names or phone numbers by email. We’ll set up a secure way to share data once we’ve agreed the mission.