Draw it
The Visual Workflow Builder lays out steps, approvals and branches on a canvas, and checks the design is runnable before you submit it.
Live now
ServiceHub
Stop grinding through repetitive back-office work. ServiceHub turns it into governed AI-agent workflows — authored visually, generated from a plain-language prompt, or started from a template — that run on demand or on a schedule you set. Human oversight goes exactly where you want it: approval gates, always-review steps, confidence thresholds and spend controls are configuration, not code. The architecture is data-driven end to end, so a new tool, a new connector or a new document format arrives as configuration rather than a release. Agents draw on 47 built-in tools and read what your systems already produce: Excel, CSV, HTML, PDF and scanned images. And every run records what the agent actually did, tool by tool — not merely what it reported. Build the next workflow yourself, today.
Three ways to build a workflow
Start wherever suits the person doing the work.
Describe it
Type what the process should do in plain language and get a working draft to refine — grounded in your own workflows, not a generic template.
Start from a template
Begin with a proven workflow and adapt it, rather than from an empty canvas.
Governance you configure, not code
The question is never "can an agent do this?" — it is "what happens when it gets something wrong?"
Human oversight as a first-class step
Approval gates and always-review steps are step types you place on the canvas, not custom code bolted on afterwards.
Confidence thresholds and branching
A step can complete on its own when it is confident and escalate when it is not. Branch conditions route the work by what the data actually says.
Spend controls
Per-workflow credit charges and balance checks stop a runaway process before it becomes an invoice.
Every run is auditable
What the agent did, not what it claimed
Each run records the tools it actually invoked and what they returned. A step that reports success without doing the work is caught, not believed.
Deliverables attached to the task
Generated workbooks, reports and posters are captured against the task record, so the evidence sits with the approval.
Reads the documents you already have
Excel, CSV, HTML, PDF and scanned images — including photographed ledgers, transcribed and reconciled without a manual re-key.
The flagship
New Business Flow Setup
Visual Workflow BuilderDescribe a business flow in plain language or build it on a visual canvas — ServiceHub reviews and runs it.
The Builder is where a new workflow is born. Author it your way: describe what you want in plain English and let the platform draft the steps, or shape it directly on a visual drag-and-drop canvas — add steps, branch on conditions, and preview any step’s real output before it ever sends. The platform checks the flow is ready to run, routes it through human review, then provisions it live. Every step can run deterministically or be handed to an AI agent, with oversight set per step from always-review to autonomous.
- Author from a prompt or on a visual drag-and-drop canvas
- Branch on conditions; preview any step’s output before it sends
- Each step: deterministic or agentic, your choice
- Human oversight per step — always-review to autonomous
Hover any step for details. Each step runs deterministically or as an AI agent.
How it works
From plain language to a governed, running workflow
- 1
Author
Describe a flow in plain language, or build it on a visual drag-and-drop canvas — branch on conditions and preview any step before it runs.
- 2
Govern
Run each step deterministic or agentic, and set human oversight per step — from always-review to fully autonomous.
- 3
Run
Every run is tool-truth verified, preview-before-send, and fully audited. Trust you can see.
Examples, not a catalogue
ReconciliationHub — the first vertical built on ServiceHub
Not a second platform — a use case assembled from the capabilities above, proven first in construction: cement and ready-mix ledgers reconciled between manufacturers, distributors and their customers, with GST and TDS handled the way Indian ledgers actually state them. The engine is verified against real counterparty workbooks at formula level, and its format training is taught and proven against your own worked example — not inferred at runtime. A new goods line or counterparty format is added as data, not a release. Growing into ReconciliationHub.
Vendor–customer ledger reconciliation
Match a vendor ledger against a customer ledger invoice-by-invoice, classify the differences (including GST and TDS), and show the arithmetic behind every figure. Matching is deterministic — a model reads a scanned page, never decides a match. The same account reconciles identically whether it arrives as a spreadsheet or a scan, checked on every build. Differences are flagged, never corrected in your data — a finance reviewer approves before the open items are emailed to the counterparty.
- Reconciliation
- GST / TDS
- Finance
A finance reviewer approves the reconciliation before any follow-up is sent — and a closing balance the platform cannot corroborate is reported as not established, never a zero that reads as agreement.
Teach it a new ledger format
A new counterparty ships a ledger the system has never seen. Upload a sample with your own description and a known-correct example — the platform learns the format, proves it against your example, pauses for your review, and an approved format reconciles from the very next run.
- Self-service
- Format training
- Reconciliation
Only a fully-proven learning auto-approves; anything less pauses for a human, who can correct it with a comment and re-run.
Execution flows teams have built
The same engine, pointed at the rest of the back office. Every workflow is customized per business — these illustrate the range, from customer feedback and scheduled outreach to document extraction and confidence-gated agent decisions, each one authored, governed and audited the same way.
Confidence-gated feedback
An agent reviews incoming feedback, acts and thanks the sender when it is confident — and escalates to a human when it is not.
- Feedback
- Confidence-based
- Auto-reply
Acts on its own above its confidence bar; hands off to a person below it.
Calendar schedule to WhatsApp
On a daily schedule, fetch a Calendly schedule, render it as a visual poster, and push it to WhatsApp.
- Calendly
- Poster image
Scheduled branded greeting
Each morning, compose a good-morning email with an inspirational quote and a branded, multilingual poster, and send it to selected customers.
- Branded poster
- Multilingual
Document extraction to report
Read student documents from Google Drive within an upload-time range, extract their data, consolidate one report, and write it back to Drive.
- Google Drive
- Extract
- Consolidated report
Answers, not a brochure
Questions people ask about ServiceHub
How work actually gets automated: authoring a workflow, deciding how much human oversight each step needs, running it, and reading the evidence afterwards.
ServiceHub turns repetitive back-office work into workflows that AI agents run under governance you configure. A workflow is a sequence of steps — some deterministic, some agentic — with human approval placed exactly where your process needs it.
The distinction that matters: the question is never "can an agent do this?" It is "what happens when it gets something wrong?" Everything else here answers the second question.
No. There are three ways in, and none of them is the lesser path — start wherever suits the person doing the work.
- Draw it
- The Visual Workflow Builder lays out steps, approvals and branches on a canvas. A readiness check verifies the design is genuinely runnable before you can submit it, so an incomplete workflow is caught at authoring time rather than halfway through its first run.
- Describe it
- State the outcome in plain language and the platform drafts the steps, forms and tool selections — grounded in your own workflows and tools, not a generic template. You refine the draft before anything is provisioned.
- Start from a template
- Begin from a workflow that already works and adapt it. Templates carry their intake form with them, so the questions come too.
Two independent dials, and keeping them separate is the point. One is how a step RUNS — deterministic or agentic. The other is how closely a human WATCHES it, at three levels. They combine freely: a step can be agentic and still require approval, or deterministic and fully autonomous. Both are configuration, not code — including the thresholds and retry limits.
- Always review
- The step pauses every time and waits for a person. Use it where being wrong is expensive.
- Confidence-based
- The step proceeds on its own when its confidence clears the bar you set, and pauses for a human when it does not. A technical failure always pauses, whatever the confidence.
- Autonomous
- The step completes without pausing. Use it where the work is routine and reversible.
- Approval checkpoints
- On top of the mode, a workflow can declare deterministic human gates that pause for an explicit approve or reject regardless of what the model decided. The agent cannot talk its way past them, because it is not consulted.
Three behaviours apply to every run, and none is optional.
- Tool-truth
- A step records the tools it actually invoked and what they returned. If it cannot prove the required tool ran, it fails to human review — never a silent false "done". This exists because the most dangerous failure in agentic software is not an error; it is a confident summary of work that never happened.
- Preview before send
- Anything with a side effect outside the platform composes a draft and holds. Nothing leaves until someone approves it.
- A full audit trail
- Reasoning, tool calls, approvals and the deliverables produced are recorded against the task, so the evidence sits with the approval instead of in a log nobody opens.
You correct it in words. A reviewer who disagrees does not have to accept the output or start over: they comment on the task in plain language — "you have mapped the tax column to the wrong header" — and re-run the step. The agent re-runs with that correction as context, and the loop repeats until it is right or the reviewer approves it directly.
Every iteration stays visible on the task. That turns review from a gate into a conversation, and it works for every agentic workflow rather than being special-cased into one.
Yes — on demand or on a schedule you set, and one workflow can carry more than one schedule.
Scheduled runs are ordinary runs: same governance, same audit trail, same pauses. A scheduled step that needs a human still waits for one. The schedule does not buy the agent extra authority.
Agents draw on 47 built-in tools and read the documents you already have — Excel, CSV, HTML, PDF and scanned images, including photographed documents, with no manual re-key.
Where a document format is genuinely new, ServiceHub can be taught it through a guided request rather than a development project: you supply a sample, describe it in your own words, and give a known-correct example the learning has to reproduce before it can go live. If it cannot reproduce your example, it does not go live — without that gate, "the system learned your format" is a claim nobody can check.
The architecture is data-driven end to end, and that is the property most worth understanding. Tool behaviour, alias maps, document dictionaries and governance thresholds are configuration rows, not compiled constants — so a new tool, a new connector or a new document format arrives as configuration rather than a release.
Connectors to external systems are registered the same way, with their credentials attached separately, so a credential can be rotated without touching the workflow that uses it.
Stated plainly, because an answer list that is all strengths is not an answer list.
- It will not act outside its governance
- A step set to always-review cannot be talked into proceeding, and a confidence gate cannot be satisfied by an agent asserting that it is confident.
- It will not silently correct your data
- Where a document is inconsistent, the platform reports the inconsistency. It does not quietly adjust figures to make an output look tidy.
- It will not claim a result it cannot evidence
- A step that cannot corroborate its output escalates rather than producing a confident answer. Going from nothing to something is the most convincing wrong answer available.