Two vendors quote the same onboarding automation and the numbers are far apart. In most cases they are not pricing the same project. One is automating a form and an email; the other is rebuilding the handoffs between sales, operations, finance and support, then proving the result works under load. The gap is scope, not markup.
This guide gives you a way to size that scope before you ask for a quote. It separates the cost components, provides an effort split you can replace with measured data, and shows the calculator logic we use when a business wants a budget range rather than a single number. No fixed BDT figure is offered here because the inputs differ too much between companies.
Define where onboarding starts and ends
Write the boundary in one sentence, for example: from a signed proposal to the customer's first successfully completed transaction. Then list every step inside that boundary. Typical steps include document collection, identity or business verification, account creation, pricing setup, credit approval, system access, welcome communication and a first-order review.
The boundary matters more than the technology. If the project ends at account creation but support still spends two weeks chasing documents, the automation has moved the work rather than removed it.
The seven cost components
- Discovery and process mapping: interviews, current-state map, exception inventory and a written scope.
- Workflow and rules build: forms, approvals, statuses, notifications, task assignment and audit history.
- Integrations: CRM, accounting, ERP, identity verification, document storage, messaging and payment systems.
- Data migration and cleanup: deduplication, field mapping, consent records and historical status correction.
- Documents and templates: contracts, welcome packs, invoices and e-signature flows.
- Testing, pilot and training: scripted cases, exception handling, role-based training and handover material.
- Running cost: subscriptions, message fees, support hours, monitoring and periodic changes.
An effort split you can calibrate
The percentages below are a starting allocation for planning, not a market benchmark. Replace them with your own estimates after discovery. Their purpose is to stop a budget from being consumed by the visible form and leaving nothing for the integration and change work.
| Phase | Planning share of build effort | What pushes it up |
|---|---|---|
| Discovery and scope | 10–15% | Multiple departments, unclear ownership, undocumented exceptions |
| Workflow and rules | 30–40% | Approval tiers, partial completion, re-entry after rejection |
| Integrations | 15–25% | No official API, unstable identifiers, two-way sync |
| Data migration | 5–15% | Duplicate records, missing fields, inconsistent status labels |
| Testing and pilot | 10–15% | High exception rate, many roles, multi-branch rollout |
| Training and handover | 5–10% | Frontline staff turnover, Bangla and English materials |
Hold contingency separately, at 10–20% of the total, and release it only against a named risk. Contingency folded into the build estimate disappears quietly.
Calculator logic for your own volumes
A useful budget model does not start with a per-customer price. It starts with complexity. Score eight inputs from one to three, where one is simple and three is complex.
| Input | Low (1) | Medium (2) | High (3) |
|---|---|---|---|
| Systems touched | One | Two or three | Four or more |
| Journey variants | One standard path | Two or three | Segment-specific paths |
| Approval levels | None or one | Two | Three or more, with thresholds |
| Documents per customer | None or one | Two or three | Four or more, with review |
| Exception share | Rare | Regular | Common and undocumented |
| Channels | Web only | Web and one messaging channel | Web, messaging, phone and branch |
| Data quality | Clean and complete | Some gaps | Duplicates and conflicting records |
| Money movement | None in onboarding | Invoicing only | Payment or credit decision |
Add the scores. A total near the bottom of the range describes a workflow project with light integration. A total near the top describes a process re-engineering programme with a software component. Use the band to choose an effort range, then multiply by your blended day rate and add licences, messaging and contingency. If you cannot estimate a band honestly, the discovery phase is underfunded.
Volume belongs in the model as a separate multiplier, not as the main driver. Twice the customers usually means more support and message cost, not twice the build effort. Peak volume matters more than average because it exposes queue and approval bottlenecks.
First-year budget worksheet
| Line | Basis | Owner | Confidence |
|---|---|---|---|
| Discovery and scope | Fixed phase, paid on sign-off | Operations lead | High |
| Workflow build | Effort band from complexity score | Product owner | Medium |
| Integrations | Per system, with a spike for unknown APIs | Technical lead | Low to medium |
| Data cleanup | Records to review × review time | Data owner | Low |
| Documents and templates | Template count × review cycles | Legal or finance | Medium |
| Testing and pilot | Scripted cases and pilot duration | QA lead | Medium |
| Training and handover | Roles × sessions, plus written guides | Operations lead | High |
| Running cost | Subscriptions, messages, support hours | Finance | Medium |
| Contingency | 10–20% of build and integration | Sponsor | Hold |
What moves the number most
- Integration surface. An official, documented API is a different project from a nightly export and a manual upload.
- Exception design. Most onboarding cost hides in the cases that do not follow the happy path.
- Data quality. Cleaning customer records is unglamorous and frequently underestimated.
- Approval policy. Every additional approver adds waiting time and a new failure state.
- Channel and language mix. WhatsApp, phone and branch intake create different evidence and consent requirements.
- AI-assisted document checks. If a model reads or classifies documents, add evaluation, human review and monitoring to both build and running cost.
Where a model is used to extract or classify documents, the governance work is part of the budget, not an afterthought. NIST's AI risk guidance and ISO/IEC 42001 describe the evaluation, documentation and monitoring expectations that a serious deployment needs. A quote that omits them is incomplete.
Measure the baseline before you automate
Record four numbers for the current process over a normal month: staff hours per onboarding, elapsed days from signature to first transaction, the share of applications that require rework, and the share of customers who drop out before completing onboarding. These become the comparison for the business case. Without a baseline, any improvement claim is a story.
Questions to ask a vendor
- Which steps are in scope, and which remain manual?
- How will you handle the exceptions we cannot describe yet?
- Which integrations have you built before, and can you show the error handling?
- What evidence proves the workflow works before go-live?
- Who owns the system after handover, and what does support cover?
- What happens to our data if we end the engagement?
Phase the spend
Fund discovery first, then a single-journey pilot, then integration and rollout. Tie each payment to an acceptance test rather than a calendar date. This keeps budget visible and makes it easier to stop a project that is not producing the operational change it promised.
Where our team fits
We design onboarding workflows, CRM connections and approval automation under Business Automation. Engagement scope and starting points are on the pricing page. If you are still choosing between a CRM, a workflow tool and a custom build, read customer onboarding automation options compared.
Sources
- McKinsey, The State of AI, accessed 17 November 2026.
- NIST AI Risk Management Framework, accessed 17 November 2026.
- ISO/IEC 42001, AI management systems, accessed 17 November 2026.
- Gartner, governance across AI agents, accessed 17 November 2026.
Responses (0)
No comments yet. Add the first one below.
Leave a response
Your email address will not be published. Required fields are marked *.