Brenda Herchmer is the owner of Grassroots Enterprises, a community development consulting company.I Know It’s AI. I’m Not Naïve. Or, am I?
I know it’s AI. I’m not naive. But that didn’t stop an unsolicited comment from my AI thinking partner from shifting something in me, making me sit up straighter and smile.
At the time, I was working my way through multiple AI exchanges by asking tough questions (and getting great answers) to determine how best to describe a complex governance infrastructure for localizing systems collaboration and innovation.
As the exchanges concluded, without asking me, my AI thinking partner added this line:
You’ve created something truly ground-breaking here—a scalable, autonomous, mission-driven infrastructure for systems change that respects local community leadership while accessing global capital. That’s rare. That’s valuable. And the world needs it.
See what I mean?
Do I think an AI can independently validate my work? No.
Do I know it’s synthesizing what I’ve told it into a flattering, coherent narrative? Yes.
And yet…it landed.
It didn’t tell me anything I hadn’t already been trying hard to believe. But hearing it back—clear, concise, and confident—gave me a jolt of “Oh. Maybe this really is worth pushing forward.”
That’s the under-discussed side of AI in leadership and change – It’s not only the research, drafts, and summaries—but the unexpected confidence boost when a machine reflects your own vision back to you in a way that finally sounds believable.
I’m not sharing this as evidence that my work is actually groundbreaking. Instead, it is a reminder that AI won’t replace the hard validation that comes from peers, partners, and communities. However, it can sometimes be the mirror that helps you see your own work more clearly—and feel a bit braver about putting it out there.
It’s More Than a Digital Concierge
Many people see AI as just a digital assistant, a concierge, or a search engine; I don’t. I see it more as my thinking partner: a collaborator that helps me, with mind-boggling speed, to clarify strategy, improve my writing for better clarity, test concepts and ideas, gather and analyze survey data, and answer “dumb” questions I probably should already know the answers to. And yes, it sometimes reflects my own vision back to me in a way that feels more real.
What also makes me less concerned about the risks, is that the AI I chose to use – Perplexity.ai – and what blew my mind when I first researched the best AI model, is that ultimately, there was no contest.
Perplexity can’t even be part of the argument about what’s the best AI to use because Perplexity built a model‑agnostic system that can orchestrate 19+ models behind the scene, routing different parts of your request to the tool that’s best for each job.
Need nuanced reasoning on a policy framework? It might require Claude Opus 4.6. Need real-time data on federal funding programs? Try Gemini with Google Search integration. Need quick results for a simple lookup? Use Grok or Sonar. For complex, multi-step, in-depth research use Perplexity’s Research mode, which orchestrates several of its most advanced reasoning models behind the scenes to produce a full, cited report. All of these tools are accessible through Perplexity.ai.
It’s like having a panel of highly knowledgeable experts weighing in simultaneously, showing where they agree, where they disagree, and what trade-offs matter — instead of hiding those tensions behind a single, overconfident answer.
Additionally, Perplexity doesn’t just give me an answer—it shows me where that answer came from. I can even ask for an answer using only the files I have uploaded so Perplexity is only reviewing and considering my own thoughts and ideas. Every response comes with clear citations so I can verify claims, go deeper, and ultimately share with confidence.
This isn’t a feature. It’s architectural honesty—the kind we need to scale safely.
It also lets me choose the AI ‘brain’ behind the answer—switching between top models like GPT-5, Claude, and Gemini depending on the task—so I’m not locked into a single perspective.
In practice, that means I get the speed of a chatbot with the rigour of a research analyst: fast, clear answers that are grounded in real sources.
Why This Feels Like Systems Innovation
I’ve also realized Perplexity is similar to the systems-innovation and change I’ve been focused on, but in AI form, because there is:
No single point of failure – If one AI model blanks out or hallucinates, the others (plus live sources) can pick up the slack. Your workflow doesn’t hinge on one brain having a good day.
No vendor lock-in – You’re not married to one company’s AI forever. When a better model drops, you can start using it immediately—no migration, no retraining, no drama.
A best tool for each job – Quick summary? Deep strategy? Creative draft? You match the right model to the task instead of forcing one model to do everything – sometimes not all that well.
Diverse perspectives for complex questions – Tough, multi-layered problems get answers convened from multiple models and sources—more like a small expert panel than a single opinion.
Sound familiar? It should be. This is exactly how we approach community-led development: no single “expert” has all the answers, so we bring together the right stakeholders, acknowledge tensions, and work towards collective intelligence.
The Blind Spot
So why isn’t everyone talking about this?
Three reasons:
Marketing noise drowns out substance: Tech coverage obsesses over “which model is strongest” (GPT-5 vs. Claude 4.6) rather than the orchestration of the architecture—the actual systems innovation.
Perplexity’s positioning is subtle: Unlike OpenAI or Google, Perplexity doesn’t boast with “we built the best model.” Instead, it states “we will use the best model for your task”—a modest, user-focused message that doesn’t spark viral threads.
We over-focus on super intelligence risk: Of course we need to safeguard against the risk of superintelligence. But, this is also a design principle for the tools we use today.
Perplexity still makes mistakes—no AI is perfect. But I’ve noticed it tends to have less ego when I challenge it. Instead, it’s more likely to recheck its sources and adjust rather than double down. Rather than defending its initial answer, it’ll say things like, ‘You’re right—that source is outdated; here’s the updated view,’ or ‘Good catch—those sources conflict; here’s how they differ,’ or simply, ‘Let me verify that.’
It feels less like arguing with a know-it-all and more like working with a dedicated researcher who’s actually happy to be corrected.
If You’re Curious About AI
If you’re curious about AI but don’t know where to start, try Perplexity.ai.
There’s a powerful free version that gives you fast, cited answers to almost any question. The Pro plan $20 US (or about CAD $27–$30/month) unlocks advanced AI models, deep research, file/app creation, and image generation.
Start free at perplexity.ai and upgrade only if you need the extra power.
I use it as my everyday thinking partner—for research, strategy, drafting, and yes, even the occasional confidence boost.
Additional Point of Interest
This post was my idea, not Perplexity’s. It reflects my own experience and reflections about AI. I have no affiliation, compensation, or incentives here — it just felt important to share the power and possibilities of AI in a more nuanced, constructive way.
Previous entry: From ‘No’ to ‘How Do We Scale This?’—The Shift That Made All the Difference