Proprietary Framework*

The Agentic AI Velocity Scale

Nine dimensions. Five stages of maturity. A structured path from first adoption to fully orchestrated AI, built for leadership teams who need clarity before committing resources.

The framework Assess your AI adoption

*Partnership credit: Joel White, Deloitte Digital Australia

Before You Begin

How to read this framework

The Velocity Scale maps AI readiness across nine dimensions and five stages of maturity. Each stage builds on the last. There are no shortcuts.

Tool agnostic. The tools listed are illustrative. The right choice depends on your stack, your contracts, and your constraints.

Responsible by design. The governance and decisioning layers exist because autonomous systems require human oversight, clear accountability, and ongoing scrutiny.

Cumulative capability. Each stage is a prerequisite for the next. You cannot orchestrate what you have not yet automated, and you cannot automate what you have not yet secured.

Human control evolves. Stages 1 to 3 keep humans in the loop. Stage 4 moves to on the loop. Stage 5, the human sets the loop. The shift is deliberate.

We use this framework to assess your current state, identify gaps, and build a sequenced programme that gets your people, systems, and leadership ready to perform with AI.

Tactical.ly

Agentic AI Velocity Scale

AI Maturity Framework

Inform

In the loop

Act

In the loop

Assist

In the loop

Automate

On the loop

Orchestrate

Sets the loop

Tooling

Chat interfaces

Standalone AI assistants. No system connections.

Claude · Microsoft Copilot

AI with tool use enabled

Chat AI connected to actions via APIs. Human-initiated.

Claude MCP · Power Automate

Agentic workflow platforms

Multi-step agents with human approval nodes.

Copilot Studio · Agentforce

Autonomous agent platforms

Self-directed pipelines. Observability essential.

AWS Bedrock Agents · Azure AI Foundry

Custom orchestration layer

Proprietary multi-agent mesh. Agents coordinate and self-direct.

Bespoke builds on hyperscaler foundations

People

Unstructured individual adoption

Self-taught, no shared practice or organisational norms.

Owner: Individual / Line manager

Centre of Excellence forms

AI CoE established. Champions embed shared habits across teams.

Owner: AI champion

Process owners + approvers

Clear accountability at every checkpoint. Structured training.

Owner: Business unit lead

AI Operations roles emerge

AIOps specialists monitor, intervene, and continuously improve agents.

Owner: COO / Transformation lead

Human-agent teaming as org design

Humans set goals and govern agent teams. Execution fully delegated.

Owner: CEO / Board

Culture

Shared awareness

Leadership names AI as strategic reality before rumour does. Clear, honest vision shared company-wide.

Owner: CEO / MD

Fear addressed directly

Managers have structured 1:1s on job security. Capability framing, not threat framing. Not a town hall.

Owner: Line managers / HR

Identity shift begins

Teams see themselves as AI-enabled, not AI-threatened. Early adopters become visible exemplars. Safety to experiment.

Owner: Team leads / People and Culture

Confidence and agency

AI use is proactive, not directed. Shared language and norms in place. Curiosity is rewarded. Fluency becomes identity.

Owner: COO

AI as cultural norm

No distinction between using AI and simply working. Embedded in onboarding and development. Leadership models it visibly.

Owner: Board / CEO

Change

Personal habit change

Low organisational effort. Individual behaviour only.

Owner: Individual / Line manager

Team workflow adoption

New norms, shared tools, peer-led change.

Owner: Team lead / AI champion

Process redesign

Approval flows, governance, retraining. BU lead and change office.

Owner: Business unit lead

Accountability restructure

Redefine performance frameworks, KPIs, and ownership models.

Owner: COO / Transformation lead

Operating model transformation

Culture, org redesign, enterprise architecture alignment.

Owner: CEO / Board

Security

Identity baseline

SSO enforced. Data classification in place. Access controls established.

Okta · Azure AD

API access controls

Scoped API keys, role-based permissions. No shared credentials.

Kong · Azure API Mgmt

Action permissions + audit logging

Every agent action logged, reviewable, and escalatable.

Datadog · Microsoft Sentinel

Autonomous failure response

Rollback capability, kill switches, anomaly detection.

CrowdStrike · PagerDuty

Agent identity management

Zero-trust. Each agent has a scoped identity. Agents verify each other.

CyberArk · HashiCorp Vault

Governance

AI acceptable use policy

Approved tools defined. Usage guidelines published. Staff acknowledgement.

Owner: General Counsel / IT

Vendor governance in place

AI vendor DPAs signed. Model training, residency, and retention covered.

Vendor DPA · Termly · Model T&Cs reviewed

Per-use-case compliance mapping

Regulatory obligations mapped. Consent mechanisms in place. AI ethics review defined.

Privacy Act · Spam Act · APRA CPS 234 · CDR

Continuous compliance monitoring

Audit-ready documentation. PIAs completed for all high-risk use cases.

Ongoing obligation tracking · PIA register

Enterprise AI governance framework

Board-level AI governance. Model risk management. AI liability standard in all contracts.

Proactive regulatory engagement · AI ethics board

Decisioning

Human decides entirely

AI generates outputs only. Human interprets and acts. No decision logic.

Owner: Individual

Rule-based logic

Explicit if/then criteria defined by humans. AI applies fixed rules, no inference.

Owner: Process owner

AI recommends, human approves

AI produces recommendation with visible reasoning. Human retains final authority.

Owner: Business unit lead

Autonomous within guardrails

Confidence thresholds set. AI decides within bounds. Edge cases escalated.

Owner: COO / AI Ops

Multi-agent consensus

Agents negotiate and arbitrate outcomes. Authority hierarchy defined by humans.

Owner: Board / CEO

Data

Session context only

You bring data to the AI in the prompt. No integration, no persistence.

Owner: Individual

Single source, read-only

AI reads from one live system via documented API. No write-back.

Snowflake · Dynamics 365

Multi-source reads

AI pulls from several systems per step. Consistent schemas required.

Fivetran · Azure Data Factory

Read and write, governed

Real-time access. AI updates records. Data quality SLAs enforced.

Collibra · Databricks Unity Catalog

Unified semantic data layer

Agents query, update, and share context in real time. Live memory.

Databricks Lakehouse · Pinecone

Integration

No integration

Standalone tools only. AI fully isolated from business systems.

Owner: Individual

Point-to-point API

Single system, stable REST endpoint, human-initiated calls.

Salesforce · ServiceNow

Managed iPaaS integration

Versioned APIs, human-gated cross-system actions.

MuleSoft Anypoint · Workato

Event-driven architecture

Async, webhook/queue-based. Full pipelines run without human initiation.

Kafka · Azure Service Bus

Agent mesh communication

Real-time inter-agent APIs. Agents invoke and coordinate autonomously.

Emerging · AWS Multi-Agent · Google A2A protocol

In the loop

On the loop

Sets the loop

See where your agentic AI maturity sits today

Nine questions, one for each dimension of the Agentic AI Velocity Scale. Pick the option that best describes your organisation as it actually works with AI today.

4 to 6 minutes

9 dimensions

Personalised heat map

A note on honesty. Pick what describes your business today. If you're unsure between two stages, go with the lower one. The shape of the result matters more than the score.

Which of these describes your business today?

Your Operating Model Scale

Here's where you sit on the velocity scale

01 Inform

02 Act

03 Assist

04 Automate

05 Orchestrate

Your overall stage

03ArchitectedCoherent

What to do next

Three moves for the most impact

Where the assessment meets the plan

Your scores, these scores are a starting point. The plan comes next.

Self-assessment is one view. A 30-minute call with Hayden tests your scores against the real operating context, identifies the two or three highest-leverage AI moves, and maps the risk surface. Most leaders leave with clearer priorities than they arrived with.

Book a 30-min call with Hayden See how agentic AI works in practice

Free. No obligation. No pitch.