AI adoption is crucial for forward thinking CFO’s but how can you budget for this?
It can be difficult to budget for the adoption of artificial intelligence (AI).
The rapid technological evolution, and the vast number of new applications that are emerging, creates deep uncertainty in cost-benefit analysis and return on investment forecasts.
CFOs have said that it can be like trying to hit an ever-moving target while blindfolded.
Two recent studies found that ROI research on AI was not always helpful, with results ranging from 5% up to 350%.
Finance leaders are increasingly concerned about the lack of clarity in the calculations, particularly given the pressure on companies to deliver AI innovations . 59% of these companies plan to increase their rollout of AI.
We’ll explore the challenges and solutions to budgeting for AI from CFOs and tech heads.
Focus on business priorities
Budgeting begins by aligning your AI project with the business goals of efficiency, growth and increased competition. It is important to avoid unnecessary spending.
Leo Smigel is the founder of algorithmic trading company Analysing Alpha. He has experienced the complexity of AI adoption. Not all AI is created equal or has the same impact on your goals.
Prioritise your tools according to the expected ROI.
Make small, agile moves…
Christoph Cemper is the founder and CEO of AIPRM prompt management platform. He says that “the field is evolving extremely fast so it’s difficult to predict what will happen in a year.”
I have found that an exploratory and iterative approach works best. It is better to take it one step at a time rather than try and forecast every application.
Starting with a small proof-of concept or pilot focused on one use is a low-risk method to collect data and build a business case for bigger investments.
Christoph suggests benchmarking AI potential against competitors that are already using AI, and not against your past performance, as AI has the potential to be disruptive.
Leo suggests an agile approach, which includes regular checks and adjustments and the use of cheap or free resources to keep costs low.
He says, “The open source ecosystem offers awesome solutions for free. Use them to avoid licensing costs.” “I recall a stock-trading AI program where we underestimated the data and training needs.
“We corrected the course using open-source software and a highly-valued solution.”
Joe Scarboro, Replan Technology’s CFO, suggests that if you want to begin exploring AI, it is best to set aside a general budget for experiments.
He says that if you do a more detailed cost-benefit analysis for each application the ROI may be outdated by the end of the exercise.
“Build a framework that is fundamental for AI experiments, and attach a budget to it. Prove how AI can be conceptually beneficial to a department or process.
Once you have this, you can get a better idea of the overall benefits and how a larger implementation might look.”
But build a long-term view
Your budget is likely to be focused on short-term objectives. Don’t overlook AI’s longer-term transformative potential.
How could you, for example:
- Enhance human capabilities throughout the organisation
- New revenue opportunities
- How can we improve our competitiveness?
Alistair Brisbourne is the head of technology for the Association of Chartered and Certified Accountants. He says that CFOs should take a longer-term view to avoid any opportunity costs.
This could include additional considerations such as scalability of the solution, employee morale and performance management. It might also consider growth opportunities or cost reduction.
It is essential that you can adapt any tool to future challenges and advancements.
Branson Knowles is the head of US Digital Banking at Top Mobile Banks. He says, “I have struggled to forecast costs and returns on AI initiatives many times.” If you adopt a holistic view, it is possible to crack the AI code.
Once you have aligned your investments with strategic goals, you will need to conduct a realistic analysis of the cost-benefit ratio. This includes everything from computing resources to employee training and solution costs.
“Be realistic about immediate costs, and keep an eye on the long-term. AI can help you shape your future.”
“Go beyond ROI forecasts to communicate AI’s powerful transformative potential.” Excel mastery is not enough. You need to have foresight.
“At first, it wasn’t easy to sell an AI virtual assistant based on the numbers. I sold the concept by showing that it would revolutionise customer service and make us look like true innovators. Years later, the AI assistant has lived up to its promise.”
Data and refinement: 2 major challenges
It is important to include data management in your AI cost-benefit analyses.
You’ll also need to budget money for data maintenance, storage, and ongoing cleaning and normalisation.
Budget for the refinements of your overall model.
Steven Kibbel is a financial planner at Day Tradingz and an editor. He says, “I have faced the unique challenges in budgeting for AI project.”
Early in my career I underestimated the need for data labelling and testing, as well as refinement. Budget overruns resulted.
“I’ve learned a lot since then.”
“Now, I suggest CFOs take into account data management, testing and refinement. Feedback, reworks, and feedback are all important. The time and resources required to gather, clean, and annotate datasets in order to train AI models are enormous.
Build generous budgets, and set realistic expectations for data operations.
Steven recommends that funds be allocated for user feedback loops, and reworks spread over a period of 12 to 24 month.
He says that AI systems must be iterated many times as they interact with users and real conditions.
It is essential to have a soft cost
You should not overlook the ‘soft costs’ of training, change management, and user adoption in your budget, as they can make or break an AI rollout.
Christoph always advocates explicitly building these costs into the budget, even if it doesn’t appear on ROI spreadsheets.
He says that failing to plan for the enabling of people and processes could derail even some of the most promising AI technologies. Soft costs are key to long-term AI success. With the right process and the right people, the rewards can be enormous.
Steven says that user adoption, personnel retraining, and workflow changes create significant costs for change management, which are often underestimated.
He says that “proper change planning is essential for ongoing budgets.”
Joe says that people are likely to perceive AI as a threat to their job. If your innovation culture doesn’t consider the human factor, even the most sophisticated budgeting and ROI analyses are doomed to failure.
AI is also a fast-moving technology, whereas organisational changes are not.
Joe proposes a phased rollout to allow people to get used to the new solution, instead of rejecting it outright.
Finding skilled AI engineers or data scientists to oversee projects is also time-consuming and expensive. For project success and budget stability, it is important to offer competitive compensation.
Consider that well-trained scientists and users will speed up adoption.
Leo says that it takes a village for AI adoption to be successful. So, mix diverse perspectives and cross-functional skills in your team.
Plan ongoing governance
There are still many companies that don’t have the right measures to deal with AI risks, including those relating to accuracy, privacy and copyright.
Understanding these risks is essential for CFOs and allocating resources to them.
Spend a portion of your budget on ongoing monitoring, assurance and governance of AI Systems to ensure compliance with existing or future regulations, ethical standards and alignment with organisational value.
Flexibility
Michael Dinich, founder of the personal finance website Wealth of Geeks and expert in AI, says that when it comes to budgeting, staying flexible is key.
He says, “I have learned to not get too attached to my projections as requirements and capabilities change faster than I expected.”
I try to make conservative estimates of initial costs and benefits. As we refine our approach and move forward with pilot projects, I will revisit these estimates frequently.
Involving stakeholders across the organisation in the budgeting is also crucial. By incorporating input from users, IT and operations, you can build confidence by identifying potential risks.
Plus, it increases support for all future changes.
Michael says that even with the best analyses, there are still uncertainties when adopting new technology like AI.
By taking a collaborative, flexible approach to budgeting and focusing my efforts on the enabling resources I was able to advance strategically.
Final thoughts on AI budgeting
Budgeting for AI projects is like navigating through a dense fog. The path keeps changing as technology advances.
The points in this article will help you to de-risk your budget and forecast potential returns.
Although there are still uncertainties, a thoughtful and systematic approach to data, testing, governance, and change management will help you improve the accuracy of your forecasts, allowing your business to reap the benefits of AI technology with minimum waste.
Download the Going for Growth report here today to find out more.
Or find out more about the power of Sage Intacct and why businesses are upgrading to a cloud-based platform.
Source: Sage
