Another day, another headline about AI! From funding announcements to new use cases, we are constantly bombarded with stories of AI’s rapid advancements. The possibilities seem endless and we are just beginning to scratch the surface.
Imagining the future of AI is exciting. However, the harder question is: What can I do now to prepare for AI adoption? Where should I start? The answer isn’t flashy or revolutionary - it’s data.
Data has always been at the heart of decision-making, from daily operations to long-term strategy. Most organizations have spent years refining data management processes to improve the efficacy of these decisions. But AI raises the stakes. Data will be the fuel that drives the power and quality of AI models deployed in our businesses. Are we ready? A good way to answer this question is to ask:
- If we fed all of our data into an AI model like ChatGPT or Gemini, would we fully trust the outputs?
- Would we be comfortable making decisions based entirely on AI recommendations?
For most of us, the honest answer is probably no – or I should say not yet.
Getting AI-Ready: Fixing the Data First
Deep learning and large language models (LLMs) excel when dealing with two kinds of data:
- Structured Data – Clearly labeled and organized datasets, such as CRM records or invoicing data. AI can efficiently analyze trends, identify patterns, and make predictions.
- Unstructured Data – Text documents, policies, emails—AI can read and extract meaning, context, and relationships from thousands of pages in seconds.
The real challenge lies in-between - where structured data is inconsistent, incomplete, or contains errors.
We’ve all looked at a dataset, spotted an anomaly, and thought, "That doesn't seem right - let’s fix it." Modern AI tools can assist with data cleansing by detecting anomalies, identifying duplicates, and suggesting data standardization. However, human oversight is still crucial to ensure accuracy.
Think of AI as an ultra-literal assistant: it doesn’t second-guess data - it believes it. Tell it that "Barry’s tea is better than Lyon’s," and it won’t question you. This is where data integrity becomes mission-critical.
Key Data Challenges to Address Before Scaling AI
"Garbage in, garbage out" (GIGO) isn’t a new concept, but AI makes the risks exponentially greater. Amazon learned this lesson the hard way in the early days when they leveraged AI models to support their hiring process. Some time later it was noted the model was biased towards male candidates leading to the project being scrapped in 2018. Instead of achieving the desired benefits – their hiring process had gone backwards.
Beyond just incorrect data, other hidden challenges can create major roadblocks:
- Manually Created Data – Many businesses rely on Excel spreadsheets, often compiled from multiple sources. While these may appear structured, they aren’t optimized for AI, leading to confusion when models process them.
- Data Hierarchy – When the same data exists in multiple places, what’s the "source of truth"? AI needs clarity on which dataset should take precedence.
- Data Definitions – Do you have a central guide defining key terms and labels? If one system defines a “customer” as an individual and another defines it as a business, AI may struggle to make sense of it.
The Path Forward: From Quick Wins toTrue Transformation
It’s tempting to jump straight to the fun stuff - flashy AI applications, productivity boosts, and automation. But the real, long-term value of AI will only come when we fix our data first. At Circit our mantra has been “fix the data, not the report”.
So, where to start?
- Audit Your Data Quality – Identify inconsistencies and gaps in structured data.
- Define a Clear Data Hierarchy – Establish a single source of truth for key datasets.
- Standardize Data Definitions – Ensure consistency across all repositories.
- Prepare Data for AI Models – Optimize structured and unstructured data for AI processing.
AI will revolutionize how we work, but only if we feed it the right information. Getting data right is the foundation - without it, even the most advanced AI won’t deliver meaningful value.
The potential is enormous, and for thosewho lay the groundwork today, the rewards will be transformational.
