The Sales Tool Graveyard in My CRM (and Where Voice AI Actually Earns Its Seat)

I once ran a count of every sales tool we’d licensed over three years. Some I’d personally championed. A handful I couldn’t remember signing up for. At least four were doing roughly the same job, and two of those nobody had logged into since onboarding week. That’s the dirty secret of the modern sales stack: it grows faster than anyone’s ability to justify it.

AI made this worse before it made it better. Every vendor slapped “AI” on their pitch deck, and suddenly a tool that reformatted your email signature was a “copilot.” So I’ve gotten allergic to the category language. What I care about now is narrow and boring: does this thing get a rep in front of a buying conversation faster, or does it just generate more activity nobody asked for?

Most AI sales tools sell you activity, not outcomes

Here’s the trap. A lot of these tools optimize for something that’s easy to measure and easy to feel good about. More emails sent. More “insights” surfaced. More tasks auto-created in your CRM. None of that is the job. The job is conversations with people who can actually buy.

I’ve watched teams light up a dashboard with green numbers while the pipeline stayed flat. The tool was working exactly as designed - it just wasn’t designed to do anything that mattered. When I evaluate something now, I ask the vendor to walk me through the path from their feature to a booked meeting or a closed deal. If they can’t draw that line without three hops and a shrug, I pass.

What actually earns a permanent seat

A short list of things I’d fight to keep. Enrichment that’s accurate enough to trust - bad data poisons everything downstream, and most enrichment is worse than people admit. Conversation intelligence that records and analyzes calls, because the gap between what reps think happened and what actually happened on a call is enormous. And anything that compresses the time between a lead raising its hand and a human talking to it.

That last one is where I’ve seen the biggest, most repeatable lift. Speed-to-lead isn’t a clever growth hack, it’s just physics. The longer a lead sits, the colder it gets, and no amount of beautifully sequenced email rescues a buyer who already talked to your competitor twenty minutes after filling out a form.

The stuff I’d skip, or at least stop pretending about

AI that writes your cold emails for you is fine in theory and mostly mediocre in practice. Buyers can smell the template. The generated paragraph that sounds “personalized” reads to a sharp prospect like exactly what it is. I’d rather have a rep write three real lines than have a machine write three hundred forgettable ones.

I’m also cool on the forecasting tools that promise to predict your quarter from CRM signals. The signals in most CRMs are too dirty for the prediction to mean much, and you end up trusting a confident number built on garbage. Fix the inputs before you buy something to interpret them. And the AI “coaching” tools that score reps on talk-to-listen ratios? Useful as a nudge, oversold as a system. A good manager listening to two calls a week beats most of them.

Where voice AI quietly belongs

For a long time I lumped “voice AI” in with the hype pile. Then I watched what happens when the phone work that nobody wants to do - and that humans do inconsistently - gets handled reliably and at volume. That changed my mind.

Think about the unglamorous, high-volume stuff. Appointment reminders. After-hours coverage when a lead comes in at 11pm. The first qualification touch on a list nobody has time to dial through. Collections follow-ups. These are conversations that have to happen, that don’t require deep relationship-building, and that human reps either rush or skip entirely when the queue gets deep. That’s the sweet spot.

At Harmony this is the part of the stack we live in - high-volume outbound and inbound calling, with the dialer, routing, and call analytics in one place so you’re not stitching three vendors together to make a phone ring intelligently. The thing I’d flag, more as an operator than a pitch: if you’re calling at real volume, deliverability and branded caller ID matter as much as the AI itself. A brilliant agent calling from a number that shows up as “Spam Likely” is just a brilliant agent talking to voicemail.

Voice AI isn’t a closer, and treating it like one backfires

Let me be clear about the boundary, because this is where I see people overreach. Voice AI is not going to run your complex, multi-stakeholder enterprise deal. It shouldn’t. The moment a conversation needs judgment, negotiation, or genuine rapport, you want a human, and you want the AI’s job to be getting that human into the room warm.

The right mental model is a tireless SDR layer plus a safety net. It works the volume so your reps spend their hours on the conversations that actually need a person. When it qualifies someone, it routes them over with context. When it can’t help, it hands off cleanly instead of trapping the caller in a loop. Get that handoff wrong and you’ve built a very expensive way to annoy people.

How I’d actually buy now

My rule of thumb after all the wasted spend: every new tool has to either replace two old ones or do something none of them could. If it just adds another tab and another login, it’s a tax, not an asset. I also kill tools on a schedule now - if a license hasn’t earned its keep in a quarter, it goes, and the team rarely notices.

Start with one painful, measurable phone problem. Leads going cold overnight, a calling queue you can’t staff, reminders that don’t get made. Run it, watch the actual conversations, count the meetings or recoveries on the other side. If you want to compare notes on where voice fits your particular mess of a stack, I’m happy to talk it through - I’ve made most of the expensive mistakes already, so you don’t have to.