Battery researchers rarely lose time because they do not understand their cells. They lose time because the answer is buried three layers below exported cycler files, copied worksheets, fragile formulas, and charts that need to be rebuilt every time a new batch finishes.
The familiar scene is not dramatic. It is Tuesday afternoon. A formation test has finished. The raw file has thousands of rows: timestamp, cycle number, voltage, current, charge capacity, discharge capacity, maybe a few temperature channels if the setup is instrumented well. The researcher knows what they want to ask:
Did this cell behave normally, and where should I look next?
Excel is usually where that question lands. Not because Excel is perfect, but because it is already in the workflow. The file is open there. The team knows how to share it. The plots for the meeting already live there. The lab notebook, the batch sheet, and the manager's questions all orbit around the spreadsheet.
The spreadsheet tax on battery research
A single cycling file can be manageable. Fifty cycles are still manageable. A month of experiments across chemistries, loadings, electrolyte variants, formation recipes, and test temperatures becomes something else.
The work starts to look like this: import the file, clean headers, check whether the cycler exported charge and discharge rows consistently, create a cycle summary table, calculate discharge capacity, calculate capacity retention, compute coulombic efficiency, make voltage curves, make a capacity fade plot, format axes, fix the chart that grabbed the wrong range, and then repeat the process when the next file arrives.
None of that is the scientific question. It is the tax paid before the scientific question can begin.
The researcher does not need another portal
Most software products ask the researcher to leave the place where the work already happens. Upload the data here. Reformat it there. Learn a separate dashboard. Ask the team to adopt another source of truth. That can work for a dedicated production analytics system, but it is heavy for the messy middle of R&D.
In battery research, the messy middle matters. A scientist may not know which comparison is important until they see the first plot. A process engineer may inspect cycle 25, then come back to cycle 21, then ask whether a temperature excursion happened before the capacity dip. The path is exploratory.
That is why an Excel-based workflow should be built around conversation inside the spreadsheet. You ask for the first view. When you feel it is right, you come back and ask the next question. The analysis does not force a new workflow just because the question changed.
What AI changes inside Excel
The useful version of AI in Excel is not a chatbot that explains how to write a formula. The useful version understands the workbook, builds the calculation, edits the sheet, and leaves the result where the team can inspect it.
A battery researcher should be able to ask:
- Summarize this cycling test and identify the main performance trends.
- Create a dashboard for voltage, capacity retention, and efficiency.
- Find cycles where discharge capacity dropped more than expected.
- Compare charge and discharge behavior across the first 50 cycles.
- Add a degradation trendline and estimate the fade rate.
The important part is that these are not isolated prompts. Each question builds on the workbook context and the previous analysis. The assistant knows which data range it used, which sheet it created, and which chart the researcher is referring to.
A representative battery cycling workflow
Imagine a 50-cycle cell test with roughly 16,000 rows of voltage, current, time, and capacity measurements. The file is not exotic. It is the kind of export battery teams deal with every week.
| Traditional step | What the researcher wants | AI-assisted interaction |
|---|---|---|
| Build formulas for each cycle | Initial capacity, final capacity, retention | "Summarize capacity by cycle." |
| Create voltage curve charts | See polarization, plateaus, and drift | "Plot voltage profiles for selected cycles." |
| Calculate efficiency manually | Catch irreversible behavior | "Calculate coulombic efficiency and flag outliers." |
| Format a meeting-ready sheet | Share the result with the team | "Create a battery performance dashboard." |
The output is not a magic answer. It is a clean workbook: a summary table, a dashboard sheet, charts that point to the right ranges, and notes that explain what changed. The researcher can inspect every number because the work remains in Excel.
Why conversation matters
Real analysis is rarely one prompt. The first answer tells you what to ask next.
A capacity fade plot might show a small drop around cycle 24. That raises a follow-up: was it charge capacity, discharge capacity, temperature, current stability, or a logging artifact? A voltage curve might look normal until you overlay only the suspect cycles. A coulombic efficiency chart might look flat until you zoom the y-axis and inspect the first five formation cycles separately.
This is where staying inside the workbook matters. The researcher does not need to export a chart, open another tool, start over, or explain the context again. They can come back when the next question becomes obvious.
Not autopilot: accountable analysis
Battery data is too consequential for a black-box assistant that silently rewrites the workbook. A useful AI assistant should show what it is about to do, execute only the requested changes, and leave a result that can be checked.
That means the assistant should be comfortable saying: here is the range I found, here are the columns I used, here is how I calculated retention, and here is the chart I created. If the workbook structure is ambiguous, it should ask for clarification instead of pretending.
Beyond batteries: the same pattern everywhere
Battery cycling data is a clear example because the files are large, the calculations are repetitive, and the interpretation matters. But the same pattern appears across technical teams.
- Quality engineers sorting through inspection reports.
- Process engineers comparing production lots.
- Finance teams assembling recurring operating reviews.
- Lab managers tracking instruments, batches, and exceptions.
- Manufacturing teams trying to separate noise from drift.
These teams do not always need a new database or a new dashboard product. Many times they need the spreadsheet they already use to become a better thinking surface.
The product principle: augment the workflow
Niobia's core product is broader than Excel. The product direction is AI process intelligence for technical work: battery R&D, manufacturing, quality, and engineering workflows where the data is fragmented and the decision requires domain context.
The principle is simple: augment the workflow before you try to replace it. Sometimes that means working inside a spreadsheet. Sometimes it means a dedicated product surface for batch review, root-cause analysis, process monitoring, or experiment interpretation.
In practice, that means:
- Stay close to the user's workflow when that is where the work already happens.
- Use natural language for intent, but keep the generated work visible.
- Let users iterate instead of forcing a perfect first prompt.
- Preserve context so follow-up questions feel natural.
- Respect the user's data, assumptions, and review process.
The best AI tools do not make experts feel like passengers. They make the routine parts of expert work lighter, so the expert can spend more time on judgment.
Where this goes next
Excel has been the default workspace for technical data for decades because it is flexible, inspectable, and already understood by almost every team. That is not going away.
What is changing is the interface. The next generation of spreadsheet work will feel less like hunting for the right formula and more like working with a careful analyst who can operate the workbook, remember the context, and wait for the next question.
For battery researchers, that means less time rebuilding the same plots and more time understanding why a cell behaved the way it did. For every other team living in Excel, the promise is similar: keep your workflow, keep your spreadsheet, and get to insight faster.
Early access
Niobia is opening early access for researchers, engineers, and analysts who spend serious time turning technical data into decisions. Battery cycling analysis in Excel is one workflow example because it is demanding, familiar, and full of repetitive work that should not slow down expert judgment.
Follow Niobia on LinkedIn for launch updates, or explore how Niobia approaches AI privacy and enterprise deployment.
FAQ
Does this replace Excel?
No. The point is to make existing technical workflows more useful. Sometimes the workbook remains the working surface. The core product is broader than Excel.
Does the researcher still review the results?
Yes. The assistant should make calculations, charts, and assumptions visible so the user can inspect and revise them.
Is this only for battery data?
No. Battery cycling is a strong starting point because the workflow is data-heavy and repetitive. The same pattern applies to quality, finance, production, and lab operations.
Why not build a separate dashboard app?
Sometimes a dedicated Niobia product surface is the right answer. But many teams already review, share, and refine parts of their work in Excel. For those workflows, bringing intelligence into the spreadsheet can be more practical than forcing a context switch.
About the author
JS is a founding engineer at Niobia, working on AI process intelligence for technical teams that need faster analysis without losing the context of how their work actually happens.