AI and data services

Our differentiator isn’t the model. It’s what sits behind it.

Human First generative AI, engineered for regulated enterprises.

We have encoded our entire delivery method into 129 governed AI agents, each carrying the recognised standards for its domain inside the prompt. An agent does not generate a plausible document – it generates our method, applied consistently, with every claim traceable to an approved source.

How we work Selected engagements

Method artefacts built as agents

Standards embedded in prompts

To draft what took 4-6 weeks

How we work

We design the accountability boundary, rather than implying it.

Generative AI ingests, synthesises and drafts at scale. The accountable expert reviews, challenges and decides. We build that boundary into the experience so it is visible to everyone who touches the output.

The agent does

Ingest, synthesise, draft

Reads hundreds of pages of requirements, standards, registers and prior decisions simultaneously. Extracts consistently, scores against a rubric and links every claim to the approved source it came from.

The expert does

Review, challenge, decide

Applies the judgement the model cannot. Interrogates what is missing rather than what is present. Owns the deliverable and the decision that follows from it, with the reasoning recorded.

The guardrails do

Constrain and evidence

Hallucination, overreliance and precedent-drift controls built into critical processes. Prompt-injection protection across ingested submissions. Control acceptance records that survive an audit and can be reviewed months after the work completed.

How an agent reaches production

Eight gates between a good idea and a governed product.

Prototypes can be built faster than most organisations can safely operationalise them. Every agent we build must pass all eight stages before it reaches production use.

01 – Discovery

Turn “we should use AI” into a named user, problem, decision and outcome.

02 – Product definition

One source of truth, so the product outlasts the meeting that produced it.

03 – Experience design

Design the accountability boundary into the product so it is visible, not merely implied.

04 – Responsible AI assessment

Governance as part of product design, not a compliance gate at the end.

05 – Cyber & architecture review

The review that decides whether this belongs in the estate at all.

06 – Identity & access validation

The stage most teams discover after launch, when their users cannot log in.

07 – UAT & validation

Tested against real operational scenarios, by the people who will be accountable.

08 – Production readiness

Deployment is not the outcome – use in a governed process is.

Selected engagements

Not pilots. Deployed, governed and in formal use.

Client identities are withheld by agreement or described by sector. Each engagement is set out in the terms the organisation measured it by.

Corrective services

Solution architecture agent, offender management

An agent embedding the StepChange Solution Architecture Method processed 400+ pages of requirements and standards in eight hours, capturing 47 non-functional requirements against a manual yield of 20-25 on comparable scope, with every element traceable to its source.

Critical infrastructure

Scope development from 200-page business case

A scope development agent converted a PAM business case into a governance-ready project scope in under four hours, surfacing 14 hidden regulated requirements and producing gap analysis across six regulatory instruments including the SOCI Act 2018 and the Essential Eight.

Regulated enterprise

AI-assisted control reviews in production

A retrieval agent synthesising evidence from approved documentation reduced the effort each recurring control review demands, with the accountable control owner retaining the final assessment. Responsible AI approved and scheduled reporting operational.

Service domains

Generative AI, data engineering and the governance around both.

Six domains, organised around the problems organisations bring us rather than by technology. Across all of them the same three commitments apply.

Human First

Generative AI ingests, synthesises and drafts at scale. The accountable expert reviews, challenges and decides. We design that boundary into the experience rather than implying it.

Method embedded

The standards are named inside the prompt — PRINCE2 processes, TOGAF phases, ISO clauses, Essential Eight controls — with self-check quality gates before anything reaches a reviewer.

Evidence traceable

Every claim links to the approved source it came from, so an output can be defended in a governance forum months after it was produced.

Six domains — select any one to see the detail.

“We have hundreds of documents holding the answer, and no way to get a consistent, defensible answer out of them at the speed we need.”

We build agents that carry a named method and its recognised standards inside the prompt, then run them over the client’s own document estate. The output is not a plausible document — it is our method applied consistently, scored against a rubric, with every claim linked to the approved source it came from.

Services
Method-embedded agents — delivery IP encoded with domain standards inside the prompt Retrieval over enterprise document estates: standards, registers, policies, prior decisions Structured extraction pipelines producing consistent, scored, evidence-linked findings Automatic traceability matrices linking every output to its originating source Prompt-injection and malicious-content protections across ingested submissions Agent tooling packaged into shared development environments for team-wide reuse
Tooling and standards
Retrieval-augmented generation over governed document sets Structured extraction and scoring rubrics Traceability matrix generation Shared development-environment extensions
Delivery Model

Outcome-based agent builds, embedded product and engineering capability, and agent enablement for teams that intend to build their own.

“There is strong interest in AI across the organisation, but the ideas are scattered, uneven in maturity and disconnected from business value or a realistic delivery path.”

We start from validated business problems rather than preferred technology. Structured discovery consolidates overlapping ideas, tests each against value, feasibility, data readiness, risk and ownership, and organises the survivors into a visible portfolio that distinguishes an early idea from a production candidate.

Services
AI opportunity discovery, use-case qualification and portfolio prioritisation Value cases grounded in validated business problems, not preferred technology Governed, time-boxed evaluation of external Gen AI platforms with firm decision dates Gen AI operating model — ownership, funding, capability and delivery pathways Build-versus-buy assessment across foundation models, copilots and bespoke agents Executive and board-level Gen AI literacy, positioning and roadmap development
Delivery Model

Strategic advisory, facilitated discovery and prioritisation programmes, and time-boxed platform evaluations with a contracted decision date.

“Prototypes are appearing faster than we can safely operationalise them, and the decisions that matter are scattered across meetings, emails and disconnected documents.”

We treat Responsible AI as part of product design. Assessments align to ISO/IEC 42001 and the NIST AI Risk Management Framework, human accountability is modelled explicitly, and control acceptance is captured as evidence a reviewer can act on. The result is a repeatable path from business problem to governed enterprise solution.

Services
Responsible AI assessments aligned to ISO/IEC 42001 and the NIST AI Risk Management Framework Human-in-the-loop accountability models with mandatory decision reasoning and explicit ownership Hallucination, overreliance and precedent-drift guardrails in critical operational processes Control acceptance records, cyber and architecture review, privacy and procurement coordination Production-readiness checks including identity, permissions and end-to-end access validation Enablement, adoption monitoring and continuity of decision rationale after go-live
Tooling and standards
ISO/IEC 42001 — AI management systems NIST AI Risk Management Framework ISO/IEC 23894 — AI risk guidance Australia’s AI Ethics Principles
Delivery Model

Independent Responsible AI assessment, embedded governance capability across an AI portfolio, and framework establishment for organisations standing up their own.

“Our data team faces growing demand without matching capacity, and the knowledge needed to move quickly is buried in fragmented technical documentation.”

Rather than adopting a single tool, we embed generative AI at each point of the data delivery lifecycle: requirements interpretation, pipeline development and debugging, documentation discovery, documentation authoring and internal tooling. The capability is then turned inward — repeatable tasks are packaged into shared tooling so individual gains become team-wide capability.

Services
AI coding assistants embedded in the development environment, co-writing and debugging against existing patterns Gen AI applied upstream to interpret dataset requirements, surfacing gaps and ambiguities before build Documentation agents over internal technical estates — conversational access to standards and runbooks Automated documentation drafted from each new pipeline script, then engineer-reviewed Custom agents packaged into shared development-environment extensions
Tooling and standards
SQL, Python and PySpark Databricks Microsoft Azure Data Snowflake ETL and ELT pipelines Data lakehouse architectures
Delivery Model

Embedded data engineering teams, uplift programmes for an existing platform team, and outcome-based delivery against a defined data scope.

“We are consolidating business-critical systems, and we cannot evidence that the data arriving on the new platform is the data that left the old one.”

Migration and integration carry the concentration of risk in any platform change. We run extraction, cleansing, validation and parallel reconciliation as a first-class workstream with its own evidence trail, and we test the integrations inside the business process rather than at the interface boundary alone.

Services
Migration frameworks covering extraction, cleansing, validation and parallel reconciliation Legacy ERP and HCM data consolidation — multi-entity, multi-country, multi-currency Integration architecture and master integration registers with end-to-end flow validation SAP data migration across S/4HANA greenfield and brownfield transformations Data reconciliation testing embedded into system integration and user acceptance cycles
Tooling and standards
SAP S/4HANA SAP BTP Integration Suite SAP XI, PI and PO MuleSoft SAP Datasphere SAP BW
Delivery Model

Outcome-based migration and reconciliation engagements, integration delivery, and specialist augmentation into an existing programme.

“We hold business-critical and regulated data, and our obligations under privacy and critical-infrastructure law are outrunning the way we actually manage it.”

We establish the enterprise and common information models, then the governance that keeps them true: stewardship and custodianship, classification, lifecycle management and metadata standards, mapped to the obligations the organisation actually carries under the Privacy Act, the Notifiable Data Breaches scheme and the SOCI Act.

Services
Enterprise and common information models within complex organisational environments Data governance policy, stewardship and custodianship frameworks Data classification, lifecycle management and metadata documentation standards Privacy Act, NDB scheme and SOCI Act compliance for business-critical data Analytics enablement — SAP Analytics Cloud, Power BI and reporting-layer design
Tooling and standards
SAP Analytics Cloud Power BI Master data management platforms Geospatial and spatial data Metadata and cataloguing tooling
Delivery Model

Governance framework establishment, embedded information architecture, and independent assessment against a named compliance obligation.

Start a conversation

Together, we turn AI potential into client performance.

Whether it is where AI belongs in your organisation, an agent you need built and governed, or a data estate that has to be understood before it can be moved – talk to the people who would do the work.

Email us +61 8 9324 1020