# Frequently asked questions

- Human page: https://www.scaledmemory.com/faq
- Markdown: https://www.scaledmemory.com/llms/faq.md
- Canonical: https://www.scaledmemory.com/faq
- Last reviewed: 2026-08-25

Direct answers about choosing an AI consultancy, scope, cost and timing, delivery and governance, and working with Scaled Memory. Each question links to its answer on the page.

## Choosing an AI consultancy

### What does an AI consultancy actually do?

An AI consultancy helps an organisation decide where AI is worth applying, then designs, builds and governs the systems that put it to work. The work usually spans three things: strategy (what to do, in what order, and why), delivery (workflows, integration and working software) and governance (controls, evaluation and accountability once AI is doing real work).

Firms weight those very differently. Some only advise. Some only build. Some hand over a report and leave. Before comparing anyone, be clear which decisions you need help making, what you expect to be left behind, and who stays accountable after the consultants leave.

Scaled Memory works across all three: we frame the decision, design the workflow and operating model around it, and equip your people to run and improve the capability.

- Answer on the page: https://www.scaledmemory.com/faq#what-does-an-ai-consultancy-do
- Explore AI Strategy, Design & Governance: https://www.scaledmemory.com/consulting

### When should an organisation bring in a specialist AI consultancy?

When the decision is consequential and internal experience has not caught up with it. Typical moments: choosing an operating model for AI across the organisation, moving a promising pilot into production, putting governance around agentic systems, or lifting capability across a workforce rather than a single team.

A useful test: if getting it wrong would cost more than the engagement — in rework, risk or lost momentum — a specialist is worth a conversation. If the need is a routine tool rollout that a vendor's standard configuration genuinely covers, you may not need a consultancy at all, and a good one will tell you so.

- Answer on the page: https://www.scaledmemory.com/faq#when-to-use-a-specialist
- Explore accelerators and products: https://www.scaledmemory.com/accelerators

### How should we compare AI consultancies in Australia?

Compare on evidence, not adjectives. Ask each firm: who will actually do the work, and what have those people delivered into production? How do you evaluate an AI system before and after release? What does the organisation own when you leave — code, prompts, documentation, patterns, trained people? How do you price, and against what outcome?

Put the same questions to every firm and keep the written answers. Treat guaranteed outcomes, unnamed “proprietary AI” and case studies that cannot be attributed or verified as warnings, whoever they come from — including us.

- Answer on the page: https://www.scaledmemory.com/faq#how-to-compare
- What to ask a consultancy to prove: https://www.scaledmemory.com/faq#what-to-ask-for-proof

### What is the difference between a boutique consultancy, a systems integrator and a Big Four firm?

They are built for different jobs. Big Four firms offer breadth, brand assurance and capacity for very large programs, priced accordingly — and the quality of your engagement depends heavily on the team you are actually assigned. Systems integrators are strongest when the job is implementing a specific platform at scale; their advice understandably tends to lead to their platforms. Boutiques put senior practitioners directly on the work, start faster and stake their reputation on each engagement, but have narrower capacity, so scope has to be honest.

Match the firm to the problem. A platform rollout suits an integrator. Board-level assurance across a very large program often suits a Big Four firm working alongside specialists. Shaping and delivering AI capability that fits how your organisation actually operates is where a senior-led boutique earns its place.

Scaled Memory is a senior-led Australian applied AI company: the people you meet are the people who do the work.

- Answer on the page: https://www.scaledmemory.com/faq#boutique-vs-integrator-vs-big-four

### What should we ask an AI consultancy to prove before appointing it?

Ask for five things: the named people who will do the work and their delivery history; a worked example or demonstration relevant to your problem, not a generic deck; the concrete method they use for evaluation and governance; references you can actually speak with; and a clear statement of what you own at the end of the engagement.

Hold us to the same standard. Scaled Memory publishes its methods on this site and provides references and worked examples in conversation. We do not yet publish named client case studies — where evidence is not yet public we say so, rather than manufacture it.

- Answer on the page: https://www.scaledmemory.com/faq#what-to-ask-for-proof

### Does Scaled Memory provide advice only, or does it build and operate systems?

Both. We help frame the decision, then design and build the workflows and systems that follow from it, and we stay accountable through adoption — governance, measurement and capability transfer included.

What we deliberately avoid is report-and-run consulting. Engagements are designed to leave behind working systems, reusable patterns and people who can run them without us.

- Answer on the page: https://www.scaledmemory.com/faq#advice-or-build
- See how we build: https://www.scaledmemory.com/ai-systems

## Scope, cost and timing

### How much does AI consulting cost in Australia?

There is no meaningful single market rate. Cost is driven by scope, the seniority of the people on the work, duration, and how much delivery — as opposed to advice — is included. Anyone quoting a universal figure is guessing.

Engagements usually take one of three shapes: a short, fixed-scope framing or strategy piece measured in weeks; a bounded first delivery or pilot measured in weeks to a few months; or ongoing delivery, governance and capability work priced against an agreed cadence. We scope and price every engagement in a written proposal before work begins, so cost is a decision you make with full information rather than a surprise.

Whoever you talk to, ask for pricing tied to a defined outcome and to the named people doing the work. That comparison exposes more than any rate card.

- Answer on the page: https://www.scaledmemory.com/faq#cost

### How long should an AI strategy, readiness or delivery engagement take?

Long enough to change something real; short enough to stay accountable. Framing and strategy work should be measured in weeks, not months. A first production workflow typically takes a small number of months, driven mostly by integration and approval overhead rather than the AI itself. Capability uplift works best as a rhythm over time rather than one long block.

Treat an engagement with no defined end point — or no working output in its first quarter — as a warning sign.

- Answer on the page: https://www.scaledmemory.com/faq#how-long

### What is the smallest useful way to start?

A focused working session with the people who own the problem: enough to frame the decision, surface the real constraints and agree whether anything further is worth doing. From there, the smallest honest commitment is a bounded first piece of work with a defined outcome — not a multi-phase program signed on day one.

- Answer on the page: https://www.scaledmemory.com/faq#smallest-start
- Start a conversation: https://www.scaledmemory.com/#contact

### How do we select and prioritise AI use cases?

Rank candidate workflows on three axes: value if it works (time, quality, risk reduction), feasibility now (data, systems access, tolerance for error) and blast radius if it misbehaves (who is affected, how visibly, how reversibly). Start where value and feasibility are high and blast radius is small.

Resist starting with the most regulated or most public workflow, however attractive its business case looks. Early deployments are how your organisation learns to evaluate, govern and trust these systems — choose ones that can safely teach you. This ranking is the core of how we frame engagements.

- Answer on the page: https://www.scaledmemory.com/faq#prioritise-use-cases
- How we frame the work: https://www.scaledmemory.com/consulting

## Delivery, governance and risk

### Why do enterprise AI pilots fail to reach production?

Because most pilots are built to demonstrate, not to operate. The recurring gaps: no integration with the systems where the work actually happens; no evaluation beyond a live demo; no owner once the innovation team moves on; no governance answer, so risk teams rightly stop it at the gate; and no adoption path for the people whose work changes.

The fix is to design for production from the first week — integration, evaluation, ownership and governance in the initial scope, not the follow-up phase. That is how we scope pilots: bounded, but built the way the production system will be built.

- Answer on the page: https://www.scaledmemory.com/faq#pilots-fail

### How do you evaluate an AI system before it is trusted with real work?

Define what “good” means with the people accountable for the work, then measure against it. That means a test set built from real, appropriately handled examples — including the awkward edge cases — assessed before launch and continuously afterwards, because behaviour drifts as models, data and usage change.

Until the evidence supports widening it, authority stays bounded: consequential steps get human review, and the system's actions are observable enough to audit. A live demo is a starting point for evaluation, never a substitute for it.

- Answer on the page: https://www.scaledmemory.com/faq#evaluate-before-trust

### How do you protect confidential data when using generative AI?

Treat it as a data governance problem before a tooling problem. Know exactly what data each workflow touches. Use enterprise agreements and deployment options where the provider commits contractually to not training on your data. Keep the most sensitive data out of prompts by design — through retrieval boundaries and redaction — rather than by asking people to be careful. Log what is sent, and review it.

For Australian organisations this sits inside existing Privacy Act and security obligations, not beside them. The goal is a workflow where the safe path is the default path, so protection does not depend on individual discipline.

- Answer on the page: https://www.scaledmemory.com/faq#confidential-data

### How should Australian organisations govern agentic AI?

Agents act, so govern them as actors rather than documents. That means bounded authority — a precise definition of what the agent may do without a person; access scoped to the task rather than inherited from a human account; evaluation and observability from the first release; human review at consequential moments; a named owner; and an escalation path for unexpected behaviour.

Governance designed into the workflow beats governance applied afterwards by committee. In regulated sectors, bring risk and compliance teams into the first scoping conversation, not the release meeting — it is the single cheapest way to avoid a stalled deployment.

- Answer on the page: https://www.scaledmemory.com/faq#govern-agentic-ai
- How we build governance in: https://www.scaledmemory.com/consulting

## Working with Scaled Memory

### Which industries does Scaled Memory work with?

Our work is designed for decision-heavy and operationally complex environments. Current areas of focus include financial services and insurance; energy, utilities and infrastructure; technology, data and product organisations; and professional and knowledge services.

These are relevant operating contexts, not a claim that every organisation in those sectors has the same needs. The practical fit depends on the workflow, the value at stake, the systems involved and who will remain accountable for the outcome.

- Answer on the page: https://www.scaledmemory.com/faq#industries
- See where we work: https://www.scaledmemory.com/#industries

### Which organisations are the best fit for Scaled Memory?

Australian organisations doing consequential work — where the workflows AI touches carry real operational, customer or regulatory weight — and leaders who want durable capability rather than a demonstration. We keep the team senior and the engagement list short, and we work directly with internal technology, risk and transformation teams rather than around them.

We are not the right fit for the cheapest possible commodity rollout, or for a report intended to sit on a shelf.

- Answer on the page: https://www.scaledmemory.com/faq#best-fit

### What happens after the first conversation?

You hear back from a senior practitioner by email — no call centre, no qualification questionnaire. If a first piece of work makes sense, you receive a short written proposal covering scope, the named people, the outcome and the price. If we are not the right fit, we say so, and where we can we point you somewhere more useful.

There is no obligation and no follow-up drip campaign. An enquiry starts a conversation, nothing more.

- Answer on the page: https://www.scaledmemory.com/faq#after-first-conversation
- Start a conversation: https://www.scaledmemory.com/#contact
