OpenAI, Google, and Microsoft are all pushing AI agents that connect to business tools and complete work on their own. This shift could change how companies run daily tasks. Instead of waiting for people to click and type, agents can act on a schedule, finish jobs, and learn from the results.
In today’s Analytics Insight Podcast, host Priya Dialani spoke with Parminder Singh, Co-founder of Redscope.ai, about why AI agents could become the new operating system of businesses.
The conversation covered his founder journey, the difference between passive software and agents, the move from growth hacking to growth engineering, and the data companies need to build AI systems that understand their business. Here are the key excerpts from the interview:
1. What led me to start Redscope.ai?
I began my entrepreneurship journey in 2014 and started Hansel.io in 2015. After I moved to the US in 2019, the company was acquired within a year. I later led Scaler’s US arm and created Redscope between 2024 and 2025. Before that, I spent about one and a half years on consulting projects across different industries. I noticed that many companies faced the same problems, so I turned some of them into products, such as LazySquid.com for SEO.
2. Why will AI agents become the operating system of businesses instead of just another software feature?
I explain it this way: traditional software is passive. It waits for a person to start every task. Agents wake up on a trigger or schedule, carry context over many days, and deliver a finished outcome. When agents succeed, companies compare them with the person they replaced, not with other software, and budget talks move between departments.
I can share an example from Tempera, Redscope’s product for direct-to-consumer brands. It builds weekly creative reports from Meta and TikTok data, and the analyst hours needed are tending toward zero. I will also be honest that 60 to 70 percent of agent deployments fail. When humans must finish the last mile, an agent feels like a feature.
3. What does the shift from growth hacking to growth engineering look like?
Growth hacking depended on a few people who held the knowledge in their heads. When they left, growth often stalled. Growth engineering takes a systems view. I have seen that marketing teams have more ideas than they can test, so the real bottleneck is the testing loop, which can take weeks. AI shortens that loop.
Our Tempera system checks a video ad 30 minutes after launch, breaks it into attributes, and finds what makes it win or lose. It then builds those winning traits into the next round. I also warn that speed cuts both ways. If the target is slightly wrong, AI drifts confidently in the wrong direction, so the hard part is judging what to optimize.
4. What data do companies need to build AI systems that understand their business?
In my view, CRMs, tickets, documents, and warehouses are not enough. I split the needs into three groups. The first is record data, which companies already have. The second is clear objectives.
Each team needs one consolidated goal, and I see this as a company problem that AI only exposes. The third is correction history, which captures the changes people make to AI outputs and why. CRMs record outcomes, not the reasoning behind decisions, and this data cannot be bought. It has to be built up inside the company.
5. How should companies start without long data-readiness projects?
I advise against spending 18 months or more preparing data while nothing ships. Companies should deploy narrow systems early, even if they are underpowered. From day one, they should record what changes people make to AI output and why. After about six months, they will hold unique context that competitors cannot buy.
Listen to the full discussion on the Analytics Insight Podcast.