Delivering worldwide · booking Q3 2026

Agents that operate your systems, not just talk about them.

Any industry, any workflow, anywhere on earth. Most AI agents can only answer questions. Ours log in, navigate, extract, reconcile and act — inside the EHRs, ERPs, claims portals, carrier websites, court filing systems, SCADA stacks and internal tools you already run. Built from scratch for your problem, deployed on your hardware, your cloud, or fully air-gapped.

Scope your agent→ Find your industry

50Industries
served

50Countries
delivered in

150Agents running
in production

agent-runtime — run RUN-2481 dry run

Illustrative trace. Every run is logged, replayable and reversible.

Any system, API or not

If a human can do it in a browser or a terminal, an agent can do it — at volume, with an audit trail.

Any industry

The hard part is never the sector. It's the legacy system underneath, and that problem is the same everywhere.

Anywhere on earth

Remote-first delivery in every timezone, with in-region data residency and local compliance.

You own the code

Full source handover on delivery. No per-seat licences, no lock-in, no black boxes.

The integration fabric

Different industry, same problem underneath

Pick a sector. The system names change, the constraints don't — no API, a closed protocol, a web-only portal, a PDF someone still prints. The agent authenticates into each one, normalises what it finds, and acts on your behalf.

Your existing system Agent traffic Healthy session Every hop authenticated, rate-limited and logged

Industries

We've built agents for these. Yours isn't on the list? Ask anyway.

Sector expertise is the fastest thing to acquire and the least of what makes an agent work. What matters is whether we can get into your systems reliably — and that skill transfers everywhere. Below, one example of what an agent does in each field.

Healthcare

Prior-authorisation agent

Pulls the clinical record, fills each payer's web-only form, tracks status and escalates denials — across a dozen payer portals with no API between them.

Financial services

KYC & onboarding agent

Gathers documents, cross-checks registries and sanctions lists, drafts the file, and routes only genuine exceptions to a compliance officer.

Insurance

Claims triage agent

Reads FNOL emails and scanned attachments, matches to the policy in a mainframe system, prices the routine ones and flags the rest.

Manufacturing

Supplier & quality agent

Watches MES and QMS signals, chases suppliers through their own portals, and files non-conformance reports with the evidence attached.

Logistics & supply chain

Carrier portal agent

Books, tracks and reconciles shipments across dozens of carrier websites that will never give you an API, and files claims on late deliveries.

Retail & e-commerce

Catalogue & pricing agent

Keeps listings synchronised across marketplaces, watches competitor pricing, and handles the returns queue end to end.

Energy & utilities

Outage & asset agent

Correlates SCADA telemetry with the asset register and the customer complaint queue to locate faults before the crew is dispatched.

Telecom

Network operations agent

Unifies provisioning, fault, monitoring and billing systems behind one interface and clears routine faults autonomously.

Public sector

Case intake agent

Digitises paper and scanned submissions, validates against the registry, and routes complete files to a caseworker — air-gapped where required.

Legal & professional

Docket & discovery agent

Monitors court e-filing systems, pulls new filings, summarises them against the matter file and drafts the internal note.

Pharma & life sciences

Regulatory dossier agent

Assembles submission packages from LIMS and document systems, checks them against the authority's format rules, and tracks the review.

Real estate & construction

Permit & compliance agent

Files applications across municipal portals, chases status, and keeps the drawing register aligned with what's been approved.

Education & research

Admissions & grants agent

Screens applications against criteria, verifies credentials with issuing bodies, and drafts funder reports from the research record.

Agriculture & food

Traceability agent

Links field data, cold-chain telemetry and buyer certification portals into a single audit trail that survives an inspection.

Travel & hospitality

Rate & distribution agent

Watches channel inventory, reconciles bookings across systems that disagree, and handles the exception queue overnight.

Media & publishing

Rights & royalty agent

Reconciles usage reports from a dozen platforms against contracts and flags underpayments with the calculation attached.

Not listed? Mining, shipping, defence, non-profit, sport, professional bodies, family offices — the question is never the industry. It's whether the work is repetitive, rule-shaped and currently done by a person clicking through a system. If that describes it, describe it to us.

Any workflow

Eight shapes of work we automate

Nearly every agent we build is one of these, or a chain of two or three. Recognising which one your problem is takes about ten minutes on a call.

Data entry & migrationMoving records between systems that will never speak to each other directly.

ReconciliationComparing two sources of truth, explaining every difference, escalating only the real ones.

Monitoring & alertingWatching a system continuously and raising an alert before a human would have noticed.

Document processingReading scans, PDFs, emails and forms, and turning them into structured, checkable records.

Research & synthesisGathering from many sources and producing a briefing with the evidence attached.

Dispatch & schedulingAssigning work against constraints, then handling the reshuffles when reality intervenes.

Quality assuranceChecking output against rules and flagging what fails, with the reason stated plainly.

ReportingAssembling the same report every period from sources that keep moving underneath you.

What we build

Five ways we deploy an agent

Same engineering discipline in every case. What changes is where it runs, what it's allowed to touch, and how hard the compliance floor is.

Our core competencyWorks in every industry

Browser workflow automation

Your team logs into a portal forty times a day to do the same six clicks. It might be a payer portal, a carrier site, a court filing system or a supplier extranet — the vendor won't build you an API, and the RPA quote came back at six months. We build an agent that does the work, with session management, retry logic, credential isolation and a full audit trail.

Integrations reverse-engineered from network traces, not guesswork
Per-tenant credential vaulting — encrypted at rest, never written to logs
Automatic re-authentication, rate-limit backoff and session recovery
Dry-run mode: see the exact diff before a single write goes live
Runs headless on your infrastructure — nothing leaves your network
Every action logged to a replayable, reversible trace

We automate systems you are authorised to access, under credentials you provide, with written authorisation on file. We will decline work that requires circumventing another party's access controls.

TIER 01

Local / Edge

Private, offline-first agents running entirely on your own hardware. No outbound calls, no third-party inference, no data leaving the building.

On-premiseOpen weightsOllama

TIER 02

Cloud native

Containerised, API-first agents in your cloud tenant. Autoscaling, queue-backed, observable, with multi-model routing and cost controls.

AWSGCPAzure

TIER 03

Enterprise

Multi-department systems with role-based access, human-in-the-loop approvals, tenant isolation, audit logging and a contractual SLA.

SSO / SAMLRBACAudit

TIER 04

Government / Defence

Air-gapped or sovereign-cloud deployment, zero-trust architecture, offline model weights, source handover and an independent security review.

Air-gappedZero trustSource escrow

Tiers & pricing

What each engagement actually includes

Indicative bands so you can qualify us before booking a call. Final scope is fixed after discovery, and we quote a firm number — not an hourly meter. We invoice in INR, USD, EUR, GBP or AED.

Deployment tiers — indicative scope and price bands
Dimension	Local / Edge	Cloud native	Enterprise	Government
Best for	Small firms, clinics, labs — anyone whose data can't leave the premises	SaaS teams, startups, D2C operations	Multi-department and regulated organisations	Ministries, defence, state utilities, public bodies
Runs on	One workstation or mini-PC. Local weights, no outbound calls	Your cloud tenant or ours. Containerised, autoscaling	Your VPC, behind your identity provider	Air-gapped or sovereign cloud. On-premise only
Core build	One agent, 2–3 tools, retrieval over your own documents	Multi-agent orchestration, API layer, monitoring, alerting	RBAC, audit log, tenant isolation, approval workflows	Zero trust, offline weights, source handover, security report
Integrations	Files, local database, one desktop application	REST / GraphQL, webhooks, message queues	ERP, CRM, EHR, ticketing — plus browser automation where no API exists	Legacy portals, closed protocols, custom hardware interfaces
Timeline	2–4 weeks	4–8 weeks	8–16 weeks	12–24 weeks, plus clearance
Indicative	₹2–6L $3–8k	₹8–25L $10–30k	₹30L–1.5Cr $40–180k	₹1Cr+ $120k+
Support	30 days of fixes included	Optional 3-month retainer	12-month SLA, 24/7 tier available	Multi-year maintenance, on-site option

← scroll the table sideways to compare tiers →

Where we differ

Why not just hire an AI agency or an RPA vendor

Both are reasonable choices for some problems. Here is where they tend to stop and we tend to start.

Capability comparison
Capability	Typical AI agency	RPA vendor	AI Newtika
Works without an API	no	partial, brittle	core competency
Runs fully offline	no	no	yes
Source code handover	sometimes	no	always
Per-seat licence lock-in	n/a	yes	none
Handles auth, sessions, rate limits	rarely	partial	built in
Dry-run before live writes	no	partial	standard
Time to first working agent	~3 months	~6 months	2–4 weeks

← scroll sideways →

Selected work

Anonymised, because most of it is under NDA

Four different industries, one underlying problem. Client names and reference contacts available on request once an NDA is in place.

Unified operations layer for a regional network operator

TELECOM · ENTERPRISE

Situation: Field and back-office staff worked across five disconnected vendor systems — provisioning, fault management, monitoring, device configuration and billing. Each had its own login, data model and quirks.
Constraint: None of the five exposed an API. The vendors had no commercial reason to build one, and replacing the stack was not on the table.
What we built: An operations layer that authenticates into each system, normalises the data into one model, and exposes a single interface — plus an autonomous fault engine that flags degradation from equipment telemetry before customers call it in.
Result: Technician workflow reduced from five system logins to one · degradation surfaced from equipment telemetry before customers report it · 88 network elements under continuous monitoring.

Browser automationMulti-tenantAnomaly detection

Prior-authorisation agent for a multi-site clinical group

HEALTHCARE · ON-PREMISE

Situation: Administrative staff spent most of each day submitting prior-authorisation requests through payer portals, then chasing status by phone and fax.
Constraint: Patient data could not leave the clinical network. Every payer had a different web-only workflow, and none offered an API.
What we built: An on-premise agent that assembles the clinical packet from the record system, submits it through each payer's own interface, polls for status, and escalates denials to a named human with the reason and the evidence attached.
Result: Packets assembled and filed without staff keying into payer portals · denials escalated to a named person with the reason and evidence attached · zero patient data leaving the clinical network · 1,000 submissions handled per week.

Local inferenceAir-gappedHuman in the loopMulti-portal

Carrier reconciliation agent for a freight forwarder

LOGISTICS · CLOUD

Situation: Operations staff tracked shipments by logging into dozens of separate carrier websites, then reconciled invoices against quoted rates by hand at month end.
Constraint: Carrier portals changed without notice, several rate-limited aggressively, and two used bot-detection that broke conventional scrapers within a week.
What we built: A resilient multi-portal agent with per-carrier session handling, adaptive backoff and structural change detection that alerts rather than silently failing — feeding a reconciliation engine that flags every billing discrepancy with the source evidence.
Result: 98 carrier portals behind one interface · portal changes raise an alert instead of failing silently · every billing discrepancy flagged with the source evidence attached.

Resilient scrapingChange detectionReconciliation

Document reconciliation agent for a compliance team

FINANCIAL SERVICES · AIR-GAPPED

Situation: A small compliance team manually cross-checked incoming documents against an internal register — thousands of records, entirely by eye, with a growing backlog.
Constraint: The documents could not leave the organisation's network under any circumstances, ruling out every hosted model.
What we built: An entirely local agent running open-weight models on a single workstation. It extracts structured fields, reconciles them against the register, and routes only genuine exceptions to a human — with the evidence for each decision attached.
Result: Runs entirely on one workstation using open-weight models · only genuine exceptions reach a human, each with its evidence attached · zero data egress · 239 documents processed per day.

Open weightsAir-gappedDocument AI

Global delivery

Wherever you are, and wherever your data has to stay

We work remote-first across every timezone, deploy into the region your regulator requires, and can travel on site for discovery and handover. Contracting, invoicing and support all adapt to where you are.

Primary hours Overlap coverage Handover notes at every shift boundary — you never re-explain

Deployment regions, residency options and the frameworks we build to
Region	Data residency	Frameworks we build to	Notes
India & South Asia	In-country cloud or fully on-premise	DPDP Act 2023, CERT-In directions, MeitY guidance	Local entity, INR invoicing, GST compliant
Europe & UK	EU or UK region, or your own datacentre	GDPR, UK GDPR, NIS2, ISO 27001 alignment	EUR / GBP invoicing, DPA and SCCs available
North America	US or Canadian region, or on-premise	HIPAA, SOC 2 alignment, PIPEDA, state privacy laws	USD invoicing, BAA available for healthcare
Middle East	In-country or sovereign cloud	UAE PDPL, Saudi PDPL, sector regulator guidance	AED / USD invoicing, on-site delivery available
Africa	In-country where available, otherwise EU region	POPIA, NDPR and national data protection acts	Offline-capable builds for low-connectivity sites
Asia-Pacific	Singapore, Australia or Japan regions	PDPA, Australian Privacy Act, APPI	Offshore contracting supported
Latin America	Brazil or US region, or on-premise	LGPD and national equivalents	Spanish and Portuguese interfaces supported

← scroll sideways →

Language. Agent interfaces, prompts and documentation delivered in the language your team works in. We've shipped in English, Hindi, Marathi, Arabic, Spanish, Portuguese, French and German — and the models we use handle far more than that.

How we work

Four stages, each with something you can hold

You get a concrete deliverable at the end of every stage. If you want to stop after discovery and take the architecture elsewhere, that's a legitimate outcome and the documents are yours.

Discovery

We map the systems, the data, the people and the failure modes. Where an integration has no API, we capture and analyse the real network traffic to establish what's actually possible.

You receive

Architecture document
Integration feasibility report
Fixed-price build quote

Build

Model selection, tool design, pipelines, orchestration. We work in short increments against your real systems in a sandbox, not against a mock.

You receive

Working agent in staging
Weekly demo, not a status email
Source access from day one

Validate

Benchmarking against a labelled evaluation set, adversarial testing, failure injection, and a dry-run period against production data with every write suppressed.

You receive

Evaluation report with pass criteria
Red-team findings and fixes
Dry-run diff for sign-off

Deploy & hand over

Deployment into your environment, monitoring and alerting wired to your on-call, and a handover session with the people who will own it after us.

You receive

Full source and runbooks
Monitoring dashboards
Support agreement

Security & data handling

The questions your security team will ask

Answered here so you don't have to schedule a call to find out. Full policy documents and a completed vendor questionnaire available on request.

Where your data lives

You choose: entirely on your hardware, inside your own cloud tenant in the region you need, or air-gapped with no network route out. We do not require a copy of your data to build.

Credential handling

Credentials are encrypted at rest with per-tenant keys, injected at runtime, and never written to logs, traces or error reports. You can rotate or revoke at any time.

Model isolation

Where a hosted model is used, we disable training on your data and document exactly which provider sees what. For sensitive tiers we use open-weight models running locally.

Audit & reversibility

Every agent action is written to an append-only trace with inputs, outputs and the decision path. Runs can be replayed, and writes can be reversed.

Authorisation

We automate only systems you are authorised to access, under your credentials, with written authorisation on file. We decline work that requires bypassing a third party's controls.

Contractual

Mutual NDA before discovery. Source code assigned to you on final payment. Source escrow and background-checked personnel available for government engagements.

Start a project

Tell us which system is the problem

Whatever your industry and wherever you are. The more specific you are about the system, the more useful our first reply will be — we answer every serious enquiry within one working day with an initial read on feasibility, not a calendar link.

Email: build@ainewtika.comReply within one working day, any timezone
Discovery call: 30 minutes, technicalBooked in your working hours, not ours
Confidentiality: Mutual NDA before discoverySend yours or use ours
Coverage: Worldwide, remote-firstOn-site available for discovery and handover

Name *

Work email *

Organisation

Country

Industry

Shape of the work

Which system should the agent work with? *

Name the actual software if you can. This single answer tells us most of what we need.

Does it have an API?

Where must the data live?

Budget range

Timeline

Who signs off?

What should the agent do?