There is no doubt we have all come across AI be it in the news, education or our day to day activities. I have and remain cautiously optimistic about the positive impacts of AI on human life. Below, i am sharing why AI Data Governance should be front and centre of AI solutions.

What is AI? AI or Artificial Intelligence is the use of computing effectively and efficiently, to replicate and automate human actions. There are 4 main types of AI however the current AI type generally used is Narrow AI which can be categorised into Generative AI, Predictive AI, Prescriptive AI and Descriptive AI. Other AI types remain theoretical concepts.

Generative AI such as ChatGPT pools data from existing information widely available to it. It uses this data to generate answers to questions. Generative AI algorithms rely on keywords to produce answers.

Predictive AI looks at what has happened in the past such as data trends, and uses the data to forecast what is expected or likely to happen in future. Most businesses include forecasting in their operations for continuity, solvency, ability to meet their obligations or market share and predictive AI can automate these operations.

Prescriptive AI looks at current problem information presented to it and compares this to identical data or scenarios that have occurred in the past, to suggest a solution to the immediate problem. Example includes diagnosis of a Medical condition or Book suggestions based on previous purchase history.

Descriptive AI looks at the metadata of existing data to provide detailed explanation or provide context to problem statements. This can be effective for example in understanding the source of an image, the types of data used to create an image to help validate its authenticity.

All of the categories of Narrow AI above require data, indeed all AI require data. Data is generated from actions and interactions. The data is then stored in a datastore where AI applications or Agents pool the data to perform the required actions in faster and efficient ways.

There is no doubt that AI is an excellent human innovation which improves processes.

However, it is important to keep sight of the fundamental drivers of a successful AI application to achieve successful business outcomes. This will help avoid issues such as hallucination of the AI application which can have significant consequences. Recall the recent expressions by Grok which caused so much media attention? One suggested reason for this was that the AI tool pooled data using the designed algorithms to provide the comments which subsequently needed to be retracted by the AI after human intervention.

Factors to consider when implementing AI solutions into your organisation include;

1. Security of your organisation’s data or proprietary information
2. Data Quality of the data being fed into the AI application
3. Data Governance – the guardrails of protection of data and prescription of how the data should be used

This list is not exhaustive.

AI Data Governance helps you define and understand where the data for your AI Solution is sourced and stored, what amount of data is required for your AI solution to prevent bias and determine if only CDEs (Critical Date Elements) are relevant to the AI solution or all data should be exposed to the AI application/Agent, as restricting data introduces bias to the AI output.

AI Data Governance provides the guardrails to ensure successful AI delivery which does not impact the organisation negatively. The founding computing principles of GIGO – Garbage in, Garbage out still applies to AI and Governance ensures this computing concept is applied to the AI solution.

All AI use a set of algorithms. An algorithm is a set of steps or a sequence of instructions to perform a task such as a cooking recipe. Algorithms are the backbone of any AI tool. Therefore, if a step is missed in the algorithm, AI reasoning becomes heavily flawed. It is therefore important to ensure that adequate governance is in place to govern the orchestration of algorithms and Machine Learning which drive the AI agent.

AI governance covers the use of AI outside an organisation ecosystem and within the ecosystem. AI data security policy is a key tool which can help organisations protect their proprietary information from being shared in the public domain by the AI agent employed. AI such as Grok, GPT etc work by pooling data from readily available sources. This can be data in various format such as text, image or audio. If a shadow AI tool which has not been commissioned by an organisation is used to carry out any form of AI activity within the organisation, this has the real potential to expose business information which could result in reputational damage and further risks to the business including its competitive advantage. It is therefore essential that governance is embedded to ensure data security is in place which defines what data is protected and needs to be made secure within the perimeter of the business.

Because AI does not do independent critical thinking but follows an algorithm predefined by a human, it is important to ensure the underlying data fed into the AI is a good and reliable representation of the organisation. Similar to existing architecture, there is the need to ensure that the stakeholders agree with the output generated by the AI solution.

Furthermore, this lack of originality by AI means better data security is now essential to protect intellectual property in the modern world.

Personal example – AI cannot build relationships with humans in a meaningful way – on a recent trip with my son to Morocco, we visited the Souk where it is customary to haggle and negotiate the price of items being purchased. I engaged a seller in what I considered intense negotiation while my son watched rather excitedly. This was done with much respect from the seller and I. I managed to secure the items for the a price I was happy with and the vendor also provided 2 cans of complimentary cold beverage to us. We became friends in that moment with promises of looking out for the seller when next we were visiting Morocco. The seller indeed came to our aid at a different stall to facilitate a sale. This is human interaction which cannot be replicated by AI. His discretion at providing us with beverage was likely due to the value he placed on the interactions with us at his stall. AI does not assess nor appreciate this emotional awareness which humans do. Whilst automating with AI should provide the same experience for every customer, it strips humans of the opportunity to thrive as a community and engage in little but powerful interactions.

Nevertheless, most businesses have functions which follow SOPs – Standard Operating Procedures using a predefined set of steps to carry out their work without individual input or discretion. AI applications provide a positive outcome for these processes in areas such as customer service where AI agents are already widely used by organisations to deliver such procedures.

Whilst it can be a frustrating experience trying to get a human customer service advisor to use empathy during interactions with their services, such expectation does not exist with AI. AI takes “Computer Says No” – a common satirical comment, to a more comforting and less frustrating level because you really know it’s the computer telling you “No” based on the data it has been fed and there is no expectation of human initiative in the decision making.

Is your organisation AI ready? AI Data governance is an area we specialise in. Reach out today to discuss your AI requirements.