Not An LLM

Human judgment for high-stakes AI decisions

You do not need a better model.
You need better judgment.

Not An LLM helps senior leaders review AI decisions before they commit budget, automate the wrong thing, or scale risk through unclear ownership.

For CTOs, CIOs, COOs, founders, and senior operators who need clear decision rights, proportionate controls, and a grounded view of where AI will actually create leverage.

Book an AI decision review See the Decision Review Sprint

For AI decisions with budget, risk, reputation, or operational consequences.

Useful before vendor selection, workflow automation, agent deployment, or board-level AI commitments.

Independent senior judgment before the organisation scales ambiguity.

Why This Matters

AI failure is often a decision failure.

Most organisations do not fail at AI because the model was not clever enough. They fail because the problem was poorly defined, the decision rights were unclear, the workflow was not understood, and accountability was spread thinly across leaders, vendors, teams, and tools.

Automation is approved before the problem is understood.

Vendor demos create confidence that the operating model cannot support.

Human-in-the-loop is mentioned, but not designed.

Teams use AI output to avoid hard trade-offs.

Tooling expands while ownership shrinks.

Agents multiply, but nobody can explain who is accountable when they act.

Leadership asks for speed without deciding what should not be automated.

The model may be capable. The system may not be ready.

AI Decision Review Sprint

AI Decision Review Sprint

A focused review of one high-stakes AI decision.

In 1-2 weeks, we examine the decision, the system around it, the people who own it, the risks it creates, and the controls needed to make it safe and useful.

The output is a clear recommendation: proceed, narrow scope, redesign, defer, buy, build, or stop.

Book an AI decision review
Duration
1-2 weeks
Price
From £7,500 + VAT
Best for
One meaningful AI decision with budget, risk, workflow, vendor, product, or accountability consequences.
Format
Interviews, decision review, system mapping, risk/control review, and executive readout.
Output
Decision memo, risk map, accountability map, recommended path, and next-step plan.

When To Use This

Use this before decisions like these.

Bring a specific decision where budget, risk, workflow, product direction, or accountability is about to move.

Should we automate this workflow?

Review whether the workflow is stable, understood, and safe enough to automate.

Should we buy this AI tool?

Test vendor claims, hidden operating costs, integration risk, governance needs, and accountability gaps.

Should we build an AI feature into our product?

Clarify user value, model risk, product liability, data exposure, and human oversight.

Should we deploy agents internally?

Assess permissions, monitoring, auditability, escalation, failure modes, and blast radius.

Should we use AI in a regulated or sensitive process?

Map risk, documentation, oversight, traceability, and decision accountability.

Why did our previous AI initiative disappoint?

Identify whether the failure was model quality, workflow design, incentives, ownership, adoption, or governance.

What You Get

What you get at the end.

You leave with a decision memo, assumption map, decision rights map, risk and constraint map, control recommendations, and a practical next-step plan.

Decision memo

A clear recommendation on whether to proceed, narrow scope, redesign, defer, buy, build, or stop.

Assumption map

The explicit and implicit assumptions behind the AI decision.

Decision rights map

Who owns the decision, who provides input, who signs off, and who carries the consequences.

Risk and constraint map

Operational, technical, data, reputational, compliance, and adoption risks.

Control design recommendations

Proportionate human oversight, escalation, monitoring, logging, and review points.

Next-step plan

A practical sequence of decisions and actions for the next 30-90 days.

Executive readout

A concise summary suitable for leadership, board, or investor conversations.

How The Review Works

Rigorous without becoming bureaucratic.

The review focuses on the decision in front of you: what is known, what is assumed, who is accountable, and what would need to be true for AI to be useful here.

Step 1

Frame the decision

We clarify the decision being made, the options available, the stakes, the people involved, and what would make the decision successful.

Step 2

Map the system

We look at workflow, incentives, data, architecture, users, vendors, governance, and failure modes as one system.

Step 3

Challenge assumptions

We test whether AI is solving the real problem or just adding capability to an unclear process.

Step 4

Design accountability

We clarify ownership, escalation, oversight, and the points where humans must remain responsible.

Step 5

Recommend the path

You receive a clear recommendation and practical next steps.

Typical Outcomes

What usually changes.

The scope narrows to where AI can create real leverage.

The organisation stops trying to automate ambiguity.

Decision rights become explicit.

Vendor claims are separated from operational reality.

Human oversight becomes designed, not assumed.

Risk controls become proportionate to the decision.

Leadership gains a clearer basis for saying yes, no, not yet, or only if.

Fit And Boundaries

Who this is for.

This is for

  • Senior leaders carrying real responsibility for AI decisions.
  • Organisations with engineers, vendors, platforms, or AI initiatives already in motion.
  • Teams that need judgment before implementation.
  • Regulated, complex, or risk-sensitive environments.
  • Leaders willing to examine incentives, ownership, and operating constraints.
  • Organisations that want to make fewer, better AI commitments.

This is not for

  • Prompt templates.
  • Chatbot builds.
  • Generic automation recipes.
  • Staff augmentation.
  • AI theatre for board slides.
  • Leaders who want certainty where the honest answer is trade-off.
  • Teams that want to delegate accountability to a model or vendor.

Pricing

Clear scope before you commit.

Start small when the decision is still forming, or use the sprint when leadership needs a defensible recommendation before budget, vendor, product, or automation commitments move further.

AI Decision Triage

£950 + VAT

90-minute working session

Best for

Best for early clarity on one live decision.

  • One live decision
  • Assumption and risk challenge
  • Short written decision note
AI Decision Review Sprint

AI Decision Review Sprint

From £7,500 + VAT

Usually £7,500-£12,500 + VAT, 1-2 weeks

Best for

Best before a budget, vendor, product, or automation commitment.

  • Decision memo
  • Risk and accountability maps
  • Control recommendations
  • Executive readout

AI Reality Check Workshop

£2,500-£3,500 + VAT

Half-day leadership workshop

Best for

Best for teams that need alignment before a larger AI initiative.

  • Leadership working session
  • Decision framing
  • Priority risks and next steps

AI Product & Governance Advisory

£3,500-£6,500/month + VAT

Ongoing senior decision support

Best for

Best for repeated AI product, governance, or adoption decisions.

  • Senior advisory cadence
  • Governance and product challenge
  • Decision support, not delivery work
Start with one real decision

Relationship To SSC

How Not An LLM relates to Sperring Software Consulting.

Not An LLM

Not An LLM is the specialist advisory brand for senior AI decisions. It starts with judgment: what should be automated, who owns the outcome, and whether the system is ready for more capability.

Sperring Software Consulting Ltd

Sperring Software Consulting Ltd is the legal entity and broader technology leadership capability. If a review shows that delivery support, architecture work, or fractional CTO help is needed, that work can continue under SSC where appropriate.

Not An LLM starts with judgment. SSC can support execution when execution is the right next step.

Visit sperring.software

About James

Technical capability meets business responsibility.

I am James Sperring, a senior technology leader and founder/operator working across product, architecture, delivery, and organisational constraints.

My work sits where technical capability meets business responsibility: what should be automated, who owns the outcome, how risk is controlled, and whether the system is ready for more capability.

Not An LLM exists because AI decisions are too often made as tooling decisions when they are really operating-model decisions.

Common Questions

Is this AI strategy consulting?

Partly, but it is narrower and more useful than a generic strategy exercise. The work starts with a specific decision, system, or commitment and tests whether the organisation is ready to make it responsibly.

Do you implement AI systems?

Not by default. Not An LLM is advisory-first. If implementation or broader delivery support is needed, that can move under Sperring Software Consulting where appropriate.

What kind of decision should we bring?

Bring a real decision with consequences: vendor selection, build-vs-buy, workflow automation, agent deployment, AI product feature, sensitive process, or a struggling AI initiative.

Will you tell us not to use AI?

Sometimes. The goal is not maximum AI adoption. The goal is better outcomes with clearer accountability.

Do we need a technical team already?

Usually yes. This is most useful for organisations with enough technical or operational complexity that AI decisions have real consequences.

Can this help with AI governance?

Yes, especially where governance needs to be practical rather than performative. The work clarifies decision rights, oversight, escalation, monitoring, and accountability.

How much does it cost?

AI Decision Review Sprint engagements start from £7,500 + VAT. Smaller decision triage sessions start from £950 + VAT. Exact scope depends on complexity, stakeholders, and risk exposure.

What happens after the review?

You may take the recommendation and execute internally. If broader technology leadership, architecture, or delivery support is needed, that can continue under SSC.

Start with one real decision.

Tell me what you are considering, what is at stake, and who owns the consequences.

Book an AI decision review

Book A Review

Start with one real decision.

Share the decision you are currently wrestling with, why it is hard, and what is at stake. If there is a fit, I will reply personally.

A useful first note names the decision, the options being considered, and who will carry the consequences.

[email protected]
LinkedIn