What Should Go Into an AI Plan and Budget?

Essentials for your AI plan and budget
December 14, 2023

AI takes a blend of both tech and teams—quickly sync up your big plans with those pesky line items.

A lot of companies were caught unawares when ChatGPT dropped into the public sphere in late 2022. Now, as budget cycles refresh, consider what your budgets for 2024 (and beyond) might need to have for your generative AI line items.

A pie chart showing a summary of the potential divisions for an AI Budget (explained in the following article)
A quick take on what an AI budget might look like. Note that purchasing 'off the shelf' products might result in a consolidation of some categories into your software licensing costs.

Many cost-pressured companies are looking for ways to save money using generative AI—including the taboo topic of staff reductions—but the reality is that initial AI spend will often be higher than meaningful savings, and it's more realistic (and arguably more ethical) to think of augmentation and re-allocation of teams rather than replacement. The budget percentages below constitute a quick take on what a non-AI-centric company, like an existing enterprise, might want to allocate for their generative AI spend to improve their existing business. In a future article, we'll discuss how to account for net new revenue and more exponential or paradigm-shifting changes to the business.

Preparation and strategy: 5%

Creation of an AI pipeline (a learning loop + innovation funnel or equivalent)

  1. Create an AI Pipeline: Set out a simple ideas funnel and learning loop to catch questions, new possibilities, and track progress with prototypes and pilots. These pipelines can be like (or integrated into) your other intrapreneurial pipelines.
  2. Host an AI Executive Forum: Rapidly upgrade thinking in the executive layer and align on the most essential imperatives and guidelines to enable exploration of new AI possibilities.
  3. Run an AI AcceleratorCreate a simple, active process for people to participate in from across your organization. This could be an immersive two days or series over twelve weeks; later, generative AI workstreams can be added to an overall accelerator or incubator function in your business.

Tip: focus on new AI use cases first, not finding a 'nail to hammer' with whatever shiny new tool is available.

Not sure where to start? Check out "Just Add Generative" for some ideas.

Talent, education and workforce: 15%

Raising AI fluency through training and practice, hiring specialists, consulting

  1. Hiring Specialists: Recruiting AI experts, data scientists, machine learning engineers, etc.
  2. Training and Development: Expenses for training existing staff in AI and machine learning concepts. AI Fluency programs should become integrated into workstreams for just-in-time learning.
  3. Consultancy Fees: If external consultants or advisors are needed.

Research and Development: 10%

Feasibility and prototyping—with potential to leverage vendor funds

  1. Feasibility Studies: Cost of preliminary research to understand the scope and potential of generative AI for your specific business needs.
  2. Prototype Development: Expenses related to developing initial versions of AI models to test concepts.

Tip: Make decisions with data throughout your AI factory process

Tech and infrastructure: 15%

Hardware (Graphic Processing Units [GPUs] for on-premise processing), software (licensing or creation) and security

  1. Hardware: Investment in servers, GPUs, or cloud services to train and run AI models.
  2. Software Licensing: Fees for AI platforms, tools, or libraries required.
  3. Security Infrastructure: Investments in cybersecurity to protect AI systems and data.

Data acquisition: 10%

Cleanup, organization, storage and/or purchasing of data to train models (or this may show up in software licensing of other organizations’ models)

  1. Data Purchasing: Costs for acquiring relevant datasets and integrating them into your data supply chain.
  2. Data Storage: Expenses for storing large volumes of data, probably in a combination of public and private clouds.
  3. Data Cleaning and Preparation: Budget for data preprocessing, which is critical for effective AI model training.

Tip: when comparing datasets to integrate into large language model training, make sure to consider the many attributes of data, such as how up to date it is.

Model development or tuning: 10%

Training, testing and validation of AI models—or may be baked into license costs

  1. Model Training: Direct costs associated with training AI models, including computing resources.
  2. Testing and Validation: Budget for rigorous testing of AI models to ensure accuracy and reliability.

Legal and compliance: 5%

Regulatory/legal preparations and alignment—may change throughout the year

  1. Regulatory Compliance: Costs associated with ensuring that AI practices comply with relevant laws and regulations.
  2. Ethical Considerations: Budget for addressing ethical issues, such as bias in AI.

Integrations: 10%

Systems integration and change management (higher for ‘exponential’ use cases than ‘incremental’ use cases, often)

  1. System Integration: Costs for integrating AI into existing IT infrastructure and business processes.
  2. Change Management: Expenses related to managing organizational changes due to AI implementation.

Maintenance and upgrades: 5%

Someone has to run it—and continue to tune models, support users, etc.

  1. Ongoing Maintenance: Regular costs for maintaining AI systems.
  2. Upgrades and Improvements: Budget for future improvements and updates to AI models and infrastructure.

Product updates: 5%

Adjustments to your products’ UX/UI, value propositions, support and documentation, if applicable

  1. Backend product changes: Impact of AI integrations into software and human workflows.
  2. Frontend UX and UI: Introduction of new interfaces such as chat or AI assistants.
  3. Support updates: Upgrading or tailoring support process and technical documentation.

Marketing and PR: 5%

If you launch AI functions, does anyone know? What if things go wrong?

  1. Promotion: Costs for marketing initiatives to promote new AI-driven products or services.
  2. Public Relations: Managing public perception and communication regarding AI initiatives.

Contingencies and new opportunities fund: 5%

There will be inevitable changes in the AI landscape

  1. Unexpected Expenses: A reserve budget for unforeseen costs or challenges in AI implementation.
  2. New Opportunities: In October 2022, almost no one had budget set aside for something like ChatGPT or other generative AI opportunities. Make sure you have enough resources to seize new opportunities—many of your competitors won't be able to.

Let us know your thoughts on AI budgets, or watch "24 Minutes on your 2024 AI Budget" on LinkedIn Live with Pam Cytron and MJ Petroni.

Watch "24 Minutes on your 2024 AI Budget" on LinkedIn Live.

Key Terms

Select any number of buttons on the left to see varieties of data sources available for analysis.