The Data Duct Tape Tax: AI Ready Data Starts with Data Strategy
5 min read
Everyone wants AI ready science. But many labs are still running on spreadsheets, disconnected systems, inconsistent metadata, and institutional knowledge trapped in people’s heads.
The result is a hidden operational cost: duplicated work, slow hand offs, unreliable data, and endless cleanup before analysis can even begin.
This is the data duct tape tax and it is one of the biggest barriers to AI readiness today.
Contents
The Cost of Disconnected Scientific Data
Raw data has potential value, but it does not automatically become usable value. We can collect endless samples, run every assay available, and sequence across every multi-omics discipline we can think of, and still struggle to extract meaningful insight if the data are disconnected, poorly described, or hard to trust.
A good business strategy might drive sales. A good scientific strategy might drive innovation. But without a good data strategy, the longevity and usefulness of all that data decays rapidly.
When data strategy is not built in parallel, “duct tape” becomes the operating model. Teams patch things together with spreadsheets, naming conventions only one person understands, and institutional knowledge living in people’s heads. Eventually the organization pays the duct tape tax: duplicated work, missing context, slow hand-offs, untrusted results, and painful cleanup at scale.
Data strategy is what keeps the science and the business from drifting apart. When it is built in parallel, teams get alignment, integrity, scale, and data that can actually support future analytics and AI. Most importantly, we turn raw data into usable value.
Building a FAIR Data Foundation
Data strategy is a plan for how data creates scientific and business value. It is a set of documented decisions around ownership, access, quality, standards, governance, and reuse.
A tech stack is part of a data strategy, but a data strategy is not a tech stack. Buying an ELN, LIMS, data warehouse, or AI product will not automatically create trusted data. Tools only create value when they support a clear strategy; otherwise, they just give teams a more expensive place to store the same confusion. That said, the right tools matter. A strong ELN implementation helps make data FAIR at the point of capture by standardizing records, organizing storage, managing access, and making information searchable
Data strategy turns scattered data into FAIR systems that support better scientific decisions, stronger operations, and more reliable analytics. In the lab, one of the most practical places to start is the electronic lab notebook, because it touches data at the moment of capture. Throughout this post, I’ll use SciNote as an example of how FAIR principles can show up in everyday scientific workflows.
Findable data can be located by the people and systems that need it. This starts with consistent naming, unique identifiers, metadata, and clear data management. People should not have to ask, “Where is that file?” or “Which version is the real one?” every time they need an answer. This is where data architecture matters. Data architecture is the city blueprint: it shows where data lives, how it connects, who owns it, and how people are expected to use it.
SciNote is an example of how findability can be built into bench workflows from the start. SciNote’s inventory management lets teams track samples, reagents, and other entities with consistent IDs and structured metadata, so a sample logged in one experiment is the same sample referenced months later. The freezer inventory map ties each digital record to a physical location, so finding a sample does not become an archaeology project.
Governance, Culture, and Data Ownership
Data governance creates alignment between people, priorities, and the data strategy plan. It keeps the business, science, and data strategy connected, and ensures policies, standards, and responsibilities are written down and followed. But governance is not policy work. It is how an organization turns data from everyone’s problem into everyone’s shared responsibility.
Good governance helps people see where their work fits in the larger data ecosystem. A data team turns standards into workflows, pipelines, and usable systems. Bench scientists are essential for data quality, metadata capture, and interpreting results. Commercial teams bring back feature requests and pain points. Leadership sets priorities and allocates resources.
A strong data governance working group brings these perspectives together. Ideally, the working group is a small group with representation from major departments. Each representative brings concerns and project needs from their area, and the group identifies patterns, recommends changes, and escalates decisions to leadership. This creates a tighter feedback loop between the people generating data, managing it, and making decisions from it.
Check out the relevant webinar:
Designing Data Strategy in the Lab: How Governance, Culture, and Connected Systems Enable Science
Building a Data Culture
Data culture is not easy to achieve, but it is essential. Here are five places to start:
- Data integrity must be non-negotiable. Every team member should feel they have time for rigorous data capture.
- Adequate resources need to exist. Resourcing and skill competency has to match the work.
- Team integration matters. Clean hand-offs, explicit lanes, and mutual respect turn departments into an ecosystem.
- AI is only as good as the data underneath it. AI improves efficiency only when the underlying data is trustworthy. The first AI mistake is implementing AI before defining the right question.
- Data literacy should be part of everyday culture. Teams should regularly share ideas, struggles, and lessons learned through workshops, lunch and learns, or any format that gets people talking across functions.
Why AI Readiness Depends on Data Strategy
Here is where most AI projects go wrong: they start with the tool. Define the problem first, then the use case, then ask whether AI is even the right answer.
Models can summarize, classify, retrieve, and generate. They cannot invent provenance you never captured, metadata you never agreed on, or decision rights you never assigned
AI accelerates pattern finding and coordination, turning hours of work into minutes. But the more work accelerates, the more important human stewardship becomes. Humans still need to define relevance, risk, quality, and action. Better AI starts with better stewardship, not bigger hype.
Key Takeaways
- Data strategy is a company strategy, not an IT project.
- Make data culture part of company culture.
- Manage data with purpose: preserve the context people need and AI systems cannot invent.
- Poor data practices slow science more than standards do.
Next Steps
- Map data flows and hand-offs: identify which steps are bottlenecks, cause friction or create pain points.
- Start with your team: align on definitions, metadata, naming conventions, sources of truth, and data pain points that slow your team down.
- Use that map to start cross team conversations: meet with the teams upstream and downstream of your work to resolve hand-off issues, clarify expectations, and reduce rework.
- Create or advocate for a data governance working group: ensure cross-functional representation.
- Choose one data quality, hand-off, pain point: develop and implement a plan to fix it at the point of capture, not at the end of the pipeline.
