01 / Build
Custom, not a chatbot
We map your real workflow first — the exceptions, the approvals, the ugly spreadsheet in the middle — then build agents against that. No generic assistant bolted onto a website.
Service proposal · AI automation · Ref AKO-AI-SVC-2026-01
We build the AI workforce
that runs inside
your operation.
Akontec designs, builds, supervises and operates virtual AI agents for any organisation — customised to your process, connected to your existing systems, and answerable to a human review desk. This site is our standing proposal: what we do, how we deliver it, and what it costs.
Free discovery call · No obligation · India and international
01 / THE PROPOSITION
Automation is easy to demo.
Automation is easy to demo.
The hard part is running it every day.
Most AI projects stall after the pilot because nobody owns them once the novelty wears off. Akontec sells the opposite: agents built to your process, put into supervised production, watched by a review desk, sampled by QC, and priced so that we only look good when they actually work.
02 / Supervise
Human-in-the-loop by design
Every agent runs at a declared autonomy level. High-risk actions are draft-only until they earn trust, and some are permanently locked behind a human — enforced in code, not left to the model.
03 / Operate
We stay on the line
Monitoring, exception handling, accuracy sampling, prompt and model refreshes, and a named delivery manager. You get an operating service, not a handover ZIP file.
02 / WHAT WE BUILD
Six service lines.
Six service lines.
One operating discipline.
Each has its own page with the full deliverable list. All six run under the same autonomy, logging and review rules described in section 03.
Service line A
Virtual AI agents
Named agents that own a function — inventory, pricing, refunds, listings, collections, scheduling — with declared autonomy and per-agent accuracy.
→ Read moreService line B
Workflow & process automation
End-to-end pipelines across your CRM, ERP, marketplace and helpdesk, with rule engines, hard guards and an audit log on every step.
→ Read moreService line C
Conversational AI
Chat, email, WhatsApp, marketplace messages and voice — AI-drafted with human approval or full auto-resolve, with handover rules per intent.
→ Read moreService line D
Document & data intelligence
Extraction, classification and validation from PDFs, scans and email, with confidence scoring and a review queue for anything doubtful.
→ Read moreService line E
AI inside your product
AI features embedded in the app, portal or site we build for you — search, recommendations, summarisation, drafting and scoring, API-first.
→ Read moreService line F
Managed AI operations
24/7 monitoring, on-call, a human review desk clearing the exception queue, weekly QC sampling and quarterly retraining.
→ Read moreReady-built solution templates — shorter build, lower cost
01 Mobility & call taxi
02 eCommerce & marketplace
03 Finance & collections
04 Real estate
05 Franchise networks
06 SaaS sales & support
07 Healthcare front office
08 Travel & holidays
Outside these eight? We still build it. A domain template typically removes two to three weeks from the build phase — nothing more. Anything genuinely new is quoted as a custom engagement.
03 / HOW YOUR AGENTS BEHAVE
Autonomy is a setting,
Autonomy is a setting,
not an accident.
Before an agent goes live, its autonomy level is written into the scope document and enforced in the platform. You always know which agents can act and which can only propose.
The agent acts, the human audits
Used where an error is reversible and cheap — inventory sync, price updates within a floor, return authorisation, forecast refresh, ad budget shifts. Actions execute immediately; a sample is re-checked by QC and every decision stays in the log.
The agent proposes, the human sends
Used where an error reaches the customer or the ledger — customer replies, listing content, refunds, compliance statements, fraud calls. Nothing leaves the building without a person clicking approve.
Hard limits
What agents may never do — enforced in code, not left to the model
- Publish medical or health claims — blocked at draft and routed to human review
- Price below the agreed margin floor — hard refusal plus an alert to the pricing owner
- Issue a refund above the approval threshold without a named human approver
- Send any reply that mentions legal action — escalated to the team lead
- Act on an unverified customer or account identity
- Move money, change bank details or alter payout instructions
- Delete records — archive and flag only
- Operate outside the log: every decision is stored with its inputs for 24 months
04 / PLANNING AID
What does your
What does your
volume look like?
A rough sizing tool using our published consumption band. Deflection rate is your assumption — we measure the real one during the supervised pilot, and we will not quote a savings percentage before we have seen your data.
Deflection means the share handled end-to-end by an agent without a human touching it. Sixty percent is a common outcome on well-defined queues; yours may be higher or much lower.
0handled by agents each month
0still routed to a human
₹0
indicative monthly consumption at ₹0.40–₹1.20 per automated task
Illustration only, not a quotation. Consumption excludes the one-time build fee and the monthly run & management plan — both published in full on the pricing page.
05 / DELIVERY MODEL
Six steps from first call
Six steps from first call
to a managed run.
Nothing starts before Schedule A is signed. Four named owners are assigned to every engagement and the escalation ladder ends at director level.
Day 0–2
Discovery call
Free. We look at the process, the volumes and the systems, and tell you honestly whether AI is the right answer.
Free
Week 1
Process map & blueprint
The workflow drawn out with exception paths, the agent list, autonomy levels and the integration inventory.
Written
Week 2
Scope lock — Schedule A
Agents, workflows, rates, volumes, thresholds and acceptance criteria signed as Schedule A. Nothing starts before this.
Binding
Week 2–6
Build & integrate
Agents built, connectors wired, guards and floors configured, dashboard stood up in your private workspace.
50% billed
Week 6–8
Supervised pilot
Everything runs draft-only against live data. We measure accuracy, tune, and promote agents to supervised one at a time.
Draft-only
Ongoing
Go live & managed run
Monitoring, exception desk, weekly QC sampling, monthly performance pack, quarterly review and retraining.
Run plan
Deemed acceptance applies seven working days after delivery of each milestone. Full delivery model, owners and escalation ladder →
06 / PRICING
Three prices, and
Three prices, and
you’ll see all of them.
Every Akontec AI engagement has the same three components: a one-time build, a monthly run & management fee, and consumption — what the models actually cost to operate. Outcome-based billing replaces the third for high-volume operations.
Component one
One-time build
Four packages from Pilot to Enterprise. Billed 50% on scope lock, 50% on acceptance of the supervised pilot.
from ₹85,000
Component two
Run & management
Three plans — Essential, Managed, Managed 24/7. Minimum term three months, then 30 days’ notice either side.
from ₹18,000 /mo
Component three
Consumption
Prepaid balance. Provider charge at cost, our 15% operations fee shown as a separate line. Hard-stop control in the dashboard.
₹0.40–1.20 /task
07 / INCLUDED IN EVERY ENGAGEMENT
No line items for
No line items for
the basics.
Eighteen things that are simply part of the engagement, whichever package you take. None of them appears as an extra on an invoice.
- Free discovery call and honest feasibility view
- Written process map and solution blueprint
- Schedule A scope lock with acceptance criteria
- Private workspace on its own environment
- Declared autonomy level for every agent
- Hard-coded guards, floors and approval thresholds
- Decision log with inputs, retained 24 months
- Pause / run control on every agent
- Live consumption ledger and hard-stop control
- Weekly QC accuracy sampling
- Exception and human review queue
- Four named owners on the account
- Five-level escalation ladder
- NDA, MSA and InfoSec baseline
- Monthly performance report
- Handover documentation and admin training
- Deemed acceptance at seven working days
- 30 days’ notice to exit a services agreement
08 / STRAIGHT TALK
What we will not promise.
Every AI vendor’s proposal has a benefits page. This is ours. Ask the same questions of whoever else you are speaking to.
100% accuracy
No model reaches it and no honest vendor claims it. We commit to a measured accuracy benchmark, weekly sampling and a review desk that catches what slips.
That AI replaces your team
It removes repetitive volume and raises what one person can carry. Where a role genuinely goes, we will say so plainly during discovery rather than after signature.
A savings percentage before we’ve seen your data
Any number quoted before we have your volumes, exception rates and system access is marketing. We quote after the process map, not before.
Results from systems we can’t reach
If access, data or a decision-maker is withheld, timelines move. We will flag it in writing the week it happens, not in the post-mortem.
Model-provider uptime
We design fallbacks, retries and degraded modes, and we monitor them. We do not own a third-party provider’s SLA and will not pretend otherwise.
Anything outside Schedule A
Verbal scope is not scope. If it matters, it goes into Schedule A before build starts — and then it is binding on us.
Do we have to change the software we already use?
No. We build around your existing stack and connect through APIs, webhooks, database views, email or file drops — whatever you already have. Replacing a system is a separate decision and we will not bundle it into an automation project to inflate scope.
Who owns the automations, the prompts and the data?
You own your data, your process logic and the configured workflows built for you. Akontec retains its own reusable platform components and domain templates. This split is written into the MSA before work starts, not argued about at exit.
What happens when the AI gets something wrong?
Three layers. High-risk actions never execute without a human. Supervised actions are sampled weekly by QC — 200 decisions re-checked at random. And every decision is logged with the inputs that produced it, so an error can be traced, corrected and guarded against rather than argued about.
How do you charge for the AI model usage itself?
As a prepaid operating balance. The provider’s charge appears at cost and our operations fee of 15% appears as a separate line. Your balance, burn rate per hour, spend today and a per-charge ledger are visible in the dashboard, and you have a hard stop control. High-volume operations can switch to outcome billing instead, where you pay per resolved unit and the model cost is our problem.
How quickly can we be live?
Fifteen days from signed scope to a first pilot agent on the Pilot package. Six weeks to a working department automation on Launch. Ten weeks for a full fleet on Operate. The clock starts at scope lock, and it depends on how fast we get system access.
Is our data used to train anyone’s models?
No. Your data is processed to perform the task and is not contributed to model training. Retention is limited to the decision log and whatever your own compliance policy requires. Data residency and on-premise deployment are available on the Enterprise package.
Next step
Send us one process. We’ll tell you what it costs to automate it.
The discovery call is free and carries no obligation. If AI is the wrong answer for your process, we would rather say so in the first hour than in month three.