I am in the middle of my company's annual planning process for our forthcoming fiscal year. It is a widely accepted corporate ritual involving myriad meetings and conferences—sales kickoffs, planning off-sites, strategy sessions, and so on—essentially a lot of pow-wows, presentations, and, of course, analysis of arrays of operational data using spreadsheets.
A standing corporate joke in the Technology Services industry I work in is that the ultimate programming language to master is spreadsheet formulas! For a large part of my career, it was Microsoft Excel, and now it happens to be Google Sheets.
What's in this article?
- What a simple shift to Google Sheets revealed about the power of AI Agents.
- Why the bigger threat to SaaS isn’t AI-generated code, but users rewiring business workflows with autonomous AI Agents.
- Why survival now depends on building context-aware Agents that understand industry nuance.
- What it takes to navigate the Agentic era and why High AI-Q™ could define who pulls ahead of the curve.
Making the transition from MS Excel to Google Sheets was a personal challenge for me.
A spreadsheet is a spreadsheet: you can add, subtract, multiply, divide, sort, and torture data in various ways to articulate why things did or did not work out the way you planned. But the subtle differences between two competing spreadsheet programs were enough to bring me to the verge of tears at times. For example, the way to filter and pick specific values in a column can be totally different, with a confusing sequence of clicks in Excel vs. Google Sheets. More on that in a different post, maybe!
Today, I want to share how the process of analyzing data on a spreadsheet has fundamentally changed and what it portends for entry-level analysts, workflows, and the Corporate SaaS business at large.
First, the Bad News
There is no sugarcoating this—and I will shed whatever little corporate doublespeak we are all guilty of at some point in our careers—AI and, in particular, co-working AI Agents, will eliminate the entry-level analysts' tasks.
There is no sugarcoating this—and I will shed whatever little corporate doublespeak we are all guilty of at some point in our careers—AI and, in particular, co-working AI Agents, will eliminate the entry-level analysts' tasks.
I had that helpful blue star of Gemini jump in every time I wanted to create a new pivot table in Google Sheets. Gemini suggested the best way to aggregate or disaggregate data, lay out the table structure, and populate the source data. There was some humility: Gemini repeatedly apologized that it is still in the process of learning, but nevertheless got it right in a fraction of the time I could do it, or I could get it done by delegating to a junior analyst in my company!
Much is being made of code written by AI Agents, but here, Gemini suggested logical and analytical workflows, which it delivered upon my permission. So, describing what I wanted done in terms of data analysis, shipping it to my Financial or Business Ops Analyst on a Monday morning, and then waiting for it to be turned around by Tuesday (hopefully with no errors), was instead completed solo by me, spending a couple of hours on the spreadsheet over the weekend! Analyst: 0, Gemini: 1
The Not-So-Good News for SaaS
One could argue that a good analyst's job comprises tasks beyond what a Manager or Executive delegates to them. These tasks include challenging hypotheses I have developed, debugging formulaic errors, and generally cleaning up the structure and cosmetics in a spreadsheet or the look-and-feel in a deck. Essentially, a human brain and an extra pair of eyes!
However, to my amazement, once it was clear what type of analysis I was trying to do (for example, "show me the mapping of my top client accounts by region and by my proprietary account stratification schema, even better, tag it to the sales person leading the account") Gemini actually started suggesting errors in my formula, diagnosed what was wrong in my source data structure etc. In other words, it started self-healing in a way beyond just debugging; it drove my analysis quickly forward and possibly even opened up new pathways in my head on how I could visualize the data!
This kind of demolished another oft-repeated corporate anodyne that AI Agents eliminate certain "tasks" but the job of an analyst would ultimately get enriched. Well, unless the analyst possibly got promoted overnight to do my job, I am not sure there is any scope for enrichment in their job, as far as this use case is concerned! Either way, it would result in one job out of the two getting eliminated!
This is what possibly triggered the “SaaSpocalypse” a couple of weeks back, wherein the real threat to SaaS/Workflow/Analysis and Visualization software was not just from users generating their own code using coding assistants but rewiring their logic, analysis, and workflows using autonomous agents.
Granted, SaaS and workflow automation tools do much more than a spreadsheet, but this is just the beginning.
SaaS and workflow automation tools do much more than a spreadsheet, but this is just the beginning.
At the click of a share button, the spreadsheet becomes a dynamic tool that can collaborate across multiple organizational personas and, in my case, connect all my colleagues around the management table, who could similarly unleash the helpful Agent to recreate customized analysis based on what I started.
Why would I need a sophisticated CRM tool other than a basic system-of-record that captures raw corporate or customer data as an input-only operation (which, by the way, could also be automated or filled in by chatbots)?
The Silver Lining—Build Agents, not LLMs or SaaS
New technologies, as experts have opined over the past few weeks, reorganize the value chain of jobs. Workers tend to migrate to higher-value jobs once the routine or “grunt work” is automated. However, that assumption is predicated on workers either being trainable to do such value-added tasks or not suffering from a generation-gap of skills, wherein younger workers take over more technologically intensive jobs. Most importantly, it is also based on the premise that the jobs being automated are routine, repetitive, and not cognitively challenging.
It looks like AI in the Agentic era is actually taking over the cognitively challenging load of tasks, so much so that some economists are encouraging governments around the world to think about a universal basic income to compensate workers who are out of a job or who no longer need to work in a society of “AI-driven abundance.” I am not going to prescribe any such strategic measures because, at one level, I lack the wisdom, and more simply put, I lack a crystal ball!
For anyone starting or in the formative stages of their career, I highly recommend focusing on Agent development.
For anyone starting or in the formative stages of their career, I highly recommend focusing on Agent development.
In my estimate, the majority of computer-science or even business professionals are shifting from building LLMs (which tends to be concentrated in 4-5 large firms) to building Agents. If one has industry domain knowledge, then that would be an added advantage, given that industry-specific contexts (for example, regulatory compliance), would be a critical aspect of Agentic workflows. And building context-aware Agents will always require human insight, because one is not just building a fancy auto-complete model, but adapting the model to the nuances of a particular organization or industry.
The process of building, testing, deploying Agents at scale, and managing them in concert with human workers is, in my opinion, the biggest opportunity facing us today.
Navigating the Shift
Navigating this shift into the Agentic era requires a level of AI maturity that we at QBurst call High AI-Q™, an enterprise’s structural readiness to deploy and manage intelligent agents at scale.
Navigating this shift into the Agentic era requires a level of AI maturity that we at QBurst call High AI-Q™, an enterprise’s structural readiness to deploy and manage intelligent agents at scale.
This is something many enterprises are only just beginning to reckon with. The transition from systems that follow a deterministic algorithm to systems that can self-heal, reason, and enforce guardrails is happening at a pace that many are struggling to catch up with. There is a lot of ground to cover before enterprises reach the level of maturity required to let these systems run truly autonomously. And it is more than just tools or models, but an organizational mindset and a technology-first ethos among the workforce.
My weekend tryst with Gemini was only possible because my data was already sitting in a clean, reasonably structured, logically-readable Spreadsheet. It is quite another task for an Agent to navigate the fragmented, messy data silos of a real-world global enterprise. Beyond the initial build, Agents must be continuously monitored and managed to ensure they scale safely and stay within the guardrails of corporate standards and ethical behavior—altogether a new engineering discipline!
We’re currently in the middle of this journey at QBurst, helping enterprises transform towards the reality of Managed Agents. Our focus is on ensuring that these Agents handle the cognitive heavy lifting while humans provide the necessary direction and oversight.
I can’t say I have all the answers. But at QBurst, we are leaning into this complexity, helping our clients and partners raise their AI quotient so that their data and structures are ready for the new era.
As Charles Dickens wrote in A Tale of Two Cities, “It was the best of times, it was the worst of times”. If "SaaSpocalypse" is on its way, as reports claim, our best chance to survive is to be the ones designing the roadmap for the hitherto uncharted new world.
I am keen to hear your views!
