You're facing conflicting views on AI in budget projections. How do you navigate the differences effectively?

To navigate conflicting views on AI budget projections, gather and understand all relevant data. Identify key disagreement areas and facilitate open discussions to hear all perspectives. Seek common ground and consult third-party experts for unbiased opinions. Develop multiple budget scenarios and assess associated risks. Propose pilot projects to test assumptions and refine projections based on real-world data. Document decisions clearly and communicate the rationale to all stakeholders to ensure alignment and transparency.

There's nothing 'artificial' about AI machine learning. AI ML forecasting is taking historical data and trends, linking it to correlative drivers in other data sets (optional) and projecting out. Demystifying how the ML models are working, how they were built, and their short comings as well as advantages will help use AI ML forecasts as a tool in budgeting & forecasting until trust can be built and the model can be trained to become the source of future budgets.

There are definitely trust issues around using AI for budgeting and forecasting. Firstly and fundamentally, there is the issue of confidentiality. If you are using a form of generative AI which learns from user input, then your data will be saved and could reappear in someone else's answer. You can reduce this risk by anonymising your input but personally, what I've read makes me unsure this is a sufficient pre-caution. Secondly, AI does make mistakes, and plenty of them, often they are just just "buried" in content which reads reasonably if reviewed in a hurry. So I would say treat AI like a junior colleague: don't give them anything too sensitive and always double-check their fundamental assumptions and spot-check their work!

I need to understand the technicality of the AI. Its key advantages, functionalities, implementation feasibility, time involves to align with my process and system. Accordingly, I will be able to judge the budget projections. Once my understanding is clear then I will be able to discuss with other team members on their projections and validate the projected budget.

Navigating conflicting views on AI in budget projections involves understanding concerns, highlighting benefits, and fostering collaboration. Identify specific issues and gather evidence to address them. Emphasize AI's accuracy, efficiency, and cost savings. Involve all stakeholders, suggest pilot programs, and offer training. Bring in experts and share successful peer experiences. Implement regular reviews, ensure transparent reporting, and address ethical concerns. Maintain clear communication and set realistic expectations to build trust and facilitate smooth AI integration.

T?o importante quanto checar o ponto de vista dos especialistas em AI é capturar as avalia??es dos especialistas da indústria / negócio / fun??o. Esses podem identificar falhas nos modelos / dados e apontar insights relevantes a serem considerados em proje??es or?amentárias que apenas a experiência prática pode proporcionar.

It's very similar to self driving cars in this respect - when there's a crash, who is accountable for a decision made by AI? I think we'll end up in a place similar to the internal controls of a business - the business wears the costs and benefits of a control failure. You may think there are no benefits to a control failure but this is not true, if you have too few people to enforce a fine, you have less payroll to pay and that is at the end of the day, a benefit. So if a business ends up solely using AI or at least under-funding their actual human professionals, they do so understanding the upside of not needing a budgeting team of 5-100 people and accepting the risk of their budgets being inhumanly inaccurate.

Much like I do in my work, identifying and either removing or at least smoothing out the data outliers is a key task. I do not think this is beyond the ability of AI to grasp, I think this concern is more apparent in a basic data model. If your AI is making this error, then it's back to the drawing board as I would expect this functionality as a minimum. I am aware this is currently a hit and miss functionality, some identify outliers and some don't. We will get to a point where outliers are easily identified and removed to address this problem of perpetual self feedback loops within AI. Another potential idea is to silo the AI's so that they have to work off the outputs of each other and critique each other, hopefully fixing these errors.

In my experience our industry does not have standard controls for the use of AI in budgeting. The only control I have seen to date is banning the use of AI altogether, which I think is smart in this experimental phase of AI but they will need to introduce it eventually and may have less capability to implement it as a result of the ban. I think developing these controls will be a key challenge for myself and others in the coming years. Ultimately we'll be the ones accountable for the AI's output. I feel developing and leading risk management in AI should be lead by either government or a professional body so we can progress it safely. Before it's too widespread to pull back if we do eventually find some form of major error.

We are just starting to comprehend what might be possible with these technologies. The earlier we embrace its potential and co-create / shape the impact and use of AI in our companies, the more we will win. All of us can win, as professionals and the more value we can create for our companies. It won’t be easy and definitely requires continuous learning for us and AI

Targeted wins first. Demonstrate that it can be done and that being uncomfortable is okay as long as we move forward and continuously create value overall.