The AntiPilot: Formulating Enterprise AI strategies and business cases to avoid languishing as perpetual AI Pilots

Happy new year!

I am working on this idea for a book with Anjali Jain . This is based on my teaching but also Erdos Research

The AntiPilot: Formulating Enterprise AI strategies and business cases to avoid languishing as perpetual AI Pilots

I got the name AntiPilot from the report last year which said that 95 percent of AI initiatives fail.

While I disagree with that report - it is true that

a) A vast majority of AI initiatives do not extend beyond the Pilot

b) and conversely, many enterprises do something simple (ex email summarisation, transcription etc) and claim that they are working with AI

To overcome this problem

1) We need a holistic perspective - bring together various parts of your enterprise

2) We need to know what questions to ask (what I call as the known unknowns)

I created this as a playbook format designed to be used in workshops

The outcome is

a) AI Product strategy

b) AI Business case

c) Enterprise learning and scaling beyond a specific product

Another way to think of the AntiPlot model is to think of how the traditional People–Process–Technology (PPT) framework evolves in the age of Enterprise AI deployments.

The traditional PPT was used in ERP rollouts, CRM implementations, government digital programmes and ITIL transformations (Information Technology Infrastructure Library) for IT Service Management (ITSM)

In the traditional PPT process definitions;

1. People: The humans who design, operate, govern, and are affected by the system. The scope includes

Scope includes:

• Roles: End-users, managers, subject-matter experts, IT staff

• Skills: Training levels, digital literacy, domain expertise

• Responsibilities: Ownership, accountability, escalation paths

• Culture: Attitudes to change, risk, compliance, learning

• Incentives: Performance metrics, reward systems, adoption drivers

Key assumption: Technology fails when people are not ready or willing to use it.

Typical questions:

• Who will use the system?

• Are they trained?

• Do they trust it?

• Who owns failures?

2. Process: The repeatable, documented workflows that define how work is done.

Scope includes:

• Business workflows: Order-to-cash, procure-to-pay, case management

• Rules & policies: Approval chains, compliance steps

• Handoffs: Who does what, when, and with what inputs

• Documentation: SOPs, swimlane diagrams, RACI matrices

• Governance: Audit trails, quality assurance, reporting

Key assumption: Automating chaos only gives you faster chaos.

Typical questions:

• What is the current workflow?

• Where are the bottlenecks?

• What must be standardised before automation?

3. Technology: The tools, platforms, and infrastructure used to support or automate the process.

Scope includes:

• Systems: ERP, CRM, HRIS, ticketing platforms

• Infrastructure: Servers, networks, storage

• Applications:Web apps, mobile apps, integrations

• Data: Databases, reporting tools

• Security: Access control, backups, disaster recovery

Key assumption: Technology is an enabler, not the solution.

Typical questions:

• What system supports this process?

• Is it reliable and scalable?

• Does it integrate with existing tools?

The Original PPT Logic is People perform Processes using Technology.

It does not cove the following cases:

• Technology changes people or

• Technology defines the process.

• Technology augments people

• Technology is not deterministic

• The process is semi automated (hence decision boundaries are not clear)

That inversion only arrives with AI.

Traditional PPT was designed for:

• Deterministic systems

• Rule-based workflows

• Predictable behaviour

• Human-led decision making

Which is why it underpins things like:

• ERP transformation

• CRM systems

There are multiple challenges in the evolution of PPT to Enterprise AI

All three must evolve together with AI. For example

1. People – Now build human intelligence around artificial intelligence.

2. Process – Turns intelligence into repeatable organisational behaviour because every team does AI differently.

3. Technology – At an enterprise level, build infrastructure that enforces learning.

Thus,

• People define what “good” looks like.

• Process encodes how intelligence is judged.

• Technology executes under constraint.

The methodology we are implementing is based on a workshop style personas - which I will discuss in the book. I have been implementing it with some organizations and also in my teaching but I am looking for organizations to pilot this (no pun intended!). Please contact me here if you are interested

Also

If you want to study#aiengineering with us, please see our course at #universityofoxford for #AI #engineering https://lnkd.in/ei6hfVAP

If you want free copies of my book on AI engineering join my substack Ajit Jaokar substack

If you want to join Erdos and be mentored by me please see the Erdos Guild initiative