Decision ledger
Every agent action written once and never rewritten: inputs, prompt, model version, tools called, citations returned, output. Append-only and hash-chained.
Append-only · hash-chainedGoogle Cloud partner · Agent governance
Anyone can ship an agent. We ship the proof.
We build agentic AI on Google Cloud, then wrap every decision in /verdict: our decision ledger, replay engine, attribution graph and policy gates. The evidence that lets autonomy run without a human watching.
- 1.4M+
- Agent decisions logged, replayable and attributed
- 82%
- Straight-through, no human in the loop
- Zero
- Unexplained decisions in production, to date
Approve claim CLM-441908 for $12,480?
Auto-approved confidence 0.94 · policy gate passed
- read policy_doc.pdf p.14 120ms
- tool claims_api.getHistory() 340ms
- ground 3 citations, 0 unsourced 210ms
- gate limit < $25k → auto 8ms
- seal hash 0xf29a… written immutable
What drove this decision
Replay on Gemini 3
Export regulator pack
Tamper-evident
Google Cloud Partner: Data & AI, Infrastructure. ISO 27001 certified · SOC 2 Type II aligned · HIPAA and GDPR delivery.
Our IP
/verdict
The proof layer for autonomous operations. Five components, one contract: every decision an agent makes is recorded, explainable, reversible and defensible.
Replay
Re-run any historical decision against a new model, a new policy or a new corpus. See what would change before you ship the upgrade.
Deterministic re-executionAttribution
Which source drove which sentence, weighted. When a regulator asks why, the answer is a document and a page number rather than a probability distribution.
Source-to-output weightingPolicy gates
Declarative thresholds that block, escalate or auto-approve at runtime. Written by your risk function, enforced in the execution path, versioned like code.
Block · escalate · auto-approveRegulator pack
One command exports the full audit trail for a date range, decision class or single case. Evidence, lineage and control attestations in a reviewable bundle.
One-command exportWhere it goes next
The same contract that proves a claims decision proves a physical one. Warehouse robotics, field service, energy dispatch, autonomous inspection. As agents move from documents to machines, the evidence requirement gets stricter, not looser.
Talk about your autonomy roadmapThe honest answer
There are hundreds of Google Cloud partners.
Here is why you would pick us.
Most of them will build you the agent. Ask what happens on the day it makes a decision nobody can explain, and the room goes quiet. That day is the whole reason we exist.
What you get
A working agent and a demo
A working agent and the evidence trail that lets it run unsupervised
When it goes wrong
Log spelunking and a best guess
Replay the exact decision, see the source that drove it, fix the gate
When risk asks
A slide deck about responsible AI
An exported regulator pack, per case, on demand
Model upgrades
Re-test by hand and hope
Replay 10,000 past decisions against the new model before cutover
What you own at the end
A statement of work and a bill
Source in your repo, running in your VPC, no lock to us
Who shows up
A pyramid, partner on the first call only
The architect who designed it stays through production
Depth on one platform, one problem owned completely. That is the trade we made instead of being available for everything.
Google Cloud practice
Deep on one platform,
rather than shallow on three
Google Cloud is where we specialize. That focus is why our estimates hold, and why our architects have already met your failure mode before you hit it.
AI & Gemini
From a grounded assistant to a fleet of agents in production, with the evaluation, cost controls and human oversight that make it deployable in a regulated setting.
Platform
- Vertex AI Platform & Model Garden
- Gemini 2.x, tuning and distillation
- Agent Builder, ADK, Agent Engine
Delivery
- Grounding on Vertex AI Search
- Eval harnesses and golden datasets
- Token economics and cost guardrails
Adoption
- Gemini Enterprise rollout
- Prompt libraries and enablement
- Human-in-the-loop review design
6 weeksPilot to production
96%Grounded answer accuracy
3.5×Analyst throughput
Data & Analytics
A BigQuery-centred lakehouse with governance that holds up under audit, and a semantic layer so AI and BI answer with the same numbers.
Foundation
- BigQuery lakehouse & BigLake
- Dataflow, Pub/Sub, Datastream CDC
- Dataform & dbt transformation
Governance
- Dataplex catalog and lineage
- Row/column policy tags, DLP
- Cost attribution by domain
Consumption
- Looker governed metrics
- BigQuery ML in-warehouse models
- Teradata, Netezza, Hadoop migration
70%Faster query performance
1 sourceOf governed truth
45%Lower warehouse TCO
Infrastructure Modernization
Landing zones, network design and migration waves, planned from a real dependency map rather than a spreadsheet of server names.
Foundation
- Terraform landing zone & org policy
- Shared VPC, Interconnect, hybrid DNS
- Billing hierarchy and FinOps setup
Migration
- Migration Center assessment & waves
- VMware estates to Google Cloud VMware Engine
- SAP, Oracle and Windows workloads
Operate
- SRE practice and SLO definition
- Committed-use discount optimization
- 24×7 managed operations
40%Run-rate reduction
ZeroUnplanned cutover downtime
99.95%Availability target
Application Modernization
Decomposition with a strangler pattern, so value ships each quarter instead of arriving in one high-risk release eighteen months out.
Runtime
- GKE Autopilot and Cloud Run
- Cloud SQL, AlloyDB, Spanner
- Apigee API management
Engineering
- Cloud Build / Cloud Deploy pipelines
- GitOps and progressive delivery
- Gemini Code Assist adoption
Quality
- Contract and load testing
- Observability with Cloud Trace
- DORA metrics baseline
70%Faster release cadence
4×Deployment frequency
60%Lower change-failure rate
Security & Responsible AI
Guardrails designed alongside the application instead of retrofitted after a review board asks how the model handles customer data.
Perimeter
- VPC Service Controls, Private Service Connect
- CMEK and Cloud KMS key strategy
- Security Command Center posture
Model safety
- Model Armor prompt/response screening
- Cloud DLP de-identification
- Red-teaming and jailbreak testing
Assurance
- Assured Workloads and data residency
- Model cards and decision logging
- Bias and fairness evaluation
100%Prompt traffic screened
0PII egress incidents
FullDecision auditability
Accelerators
The build layer,
on top of the proof layer
Reusable engineering that removes the first eight weeks from a build. Each one is instrumented for /verdict out of the box, and each ships as source you own, deployed inside your own project and VPC.
groundwork
GA
A grounded-assistant framework on Gemini and Vertex AI Search: ingestion, chunking, citation rendering, eval harness and cost telemetry, ready for your corpus.
- Deploys in
- 2 weeks
- Runs on
- Cloud Run · Vertex AI
- Best for
- Knowledge & support
docflow
GA
Document processing pipeline for claims, KYC and lending. Document AI extraction with Gemini fallback, confidence routing and a human review queue.
- Deploys in
- 3 weeks
- Runs on
- Document AI · Workflows
- Best for
- Insurance & banking
lakebase
GA
Opinionated BigQuery foundation: medallion layout, Dataform models, Dataplex governance, policy tags and cost dashboards as Terraform from commit one.
- Deploys in
- 2 weeks
- Runs on
- BigQuery · Dataplex
- Best for
- Data platform resets
verdict
Our IP
The proof layer itself: decision ledger, replay engine, attribution graph, policy gates and regulator-pack export. Wraps any agent runtime, including ones we did not build.
- Deploys in
- 4 weeks
- Runs on
- Agent Builder · GKE · BigQuery
- Best for
- Regulated autonomy
Industries
Domain context
as well as cloud skills
We bring reference data models, regulatory constraints and the failure modes specific to your sector, so discovery starts from something rather than a blank page.
- Financial ServicesRisk analytics, payments intelligence, capital-markets research
- InsuranceUnderwriting assistance, claims automation, bordereaux ingestion
- HealthcareClinical documentation, prior authorization, patient access
- Life SciencesTrial data harmonization, pharmacovigilance, medical affairs
- Retail & CPGDemand forecasting, merchandising, loyalty personalization
- ManufacturingPredictive maintenance, visual quality, supply visibility
- Media & EntertainmentMetadata enrichment, content supply chain, audience models
- Public SectorCitizen services, records digitization, open data platforms
Client work
Outcomes we
can point at
Selected engagements across financial services, insurance, healthcare and retail, with the numbers that mattered to the sponsor.
Loading client work…
Company
The bottleneck on autonomy was never the model. It is that nobody could prove why the machine decided what it decided.
Jaspreet Kaur, Founder & CEO, on why CloudInclusive exists
Jaspreet Kaur
Founder & CEO
Entrepreneur and business leader with over a decade across business analysis, cloud infrastructure and data analytics. Jaspreet founded CloudInclusive in 2021 to close the gap between the slide deck and the running system.
Under her leadership the firm has guided 20+ customers through cloud and AI transformation. She has worked with SAP and Fourth Quadrant Labs, is an active member of the Entrepreneurs' Forum, and advocates for women building technology businesses.
Expertise
- Cloud infrastructure & deployment
- Business analysis & strategy
- Data analytics & data mining
- Talent management & operations
Education
- MBA, Business AnalyticsBITS Pilani · 2020-2022
- MSc, BiotechnologyBarkatullah University · 2009-2011
10+ years experience
20+ customers guided
Founded 2021
Bangalore, India
Latest insights
We publish what
we learned the hard way
Writing on agent governance, evidence design and the economics of proof, written by the engineers doing the work rather than a content team.
Loading insights…
Life at CloudInclusive
Build AI that ships
Small teams, real production systems, direct access to the client's problem. No bench, no shadow work.
Loading open roles…
Nothing matching your discipline? Send your work to careers@cloudinclusive.in.
Next step
Prove one decision.
Then automate the rest.
Two weeks. We take one real decision your business makes today, wire it through /verdict, and hand you the ledger, the replay and the business case. Yours to keep, whatever you decide next.
- Response within one business day
- Senior architect on the first call
- No obligation to proceed