What we learned from 200 customer interviews

We've done somewhere north of 200 customer interviews since we started Avery NXR. Some structured, some casual. Some with current customers, some with prospects, some with people who decided not to buy. Some recorded, some just notes.

We're going to share what we've learned across them. Not just product insights — broader patterns about how AI agent buyers actually think.

What buyers say they want vs. what they actually buy

The biggest gap in our interviews: what people say they want and what they actually pay for are often different.

What buyers SAY they want when asked open-ended: → Sophisticated multi-agent orchestration → Cutting-edge models → Beautiful interfaces → Lots of features → Integration with everything

What buyers actually pay for when watched closely: → Specific workflows that absorb specific pain → Reliable execution on what they've already configured → Templates that work out of the box → Clear pricing → Few specific integrations they care about

The aspirational features get into demos. The actual features get into invoices.

The deciding factor varies by segment

Different customer segments make decisions on different criteria:

Solo operators / freelancers: → Cost is the dominant factor → "Will this pay back for me, this year, with my volume?" → Free Desktop tier is what gets them in the door → Pro tier only if specific need emerges

5-30 person teams: → Time-to-value matters most → "Can I get this running in a week and feel results immediately?" → Templates are crucial because building from scratch is too slow → Champion-led adoption (one person evangelizes)

30-200 person teams: → Total cost projection is the deciding factor → "What does this cost when we have 50 users?" → Flat per-user pricing wins because predictable → IT/security review is a real gate

200+ person enterprises: → Architectural fit + compliance matters most → "Does this work with our existing infrastructure + governance?" → Data residency / audit transparency are dealbreakers → Vendor longevity matters (will this company exist in 3 years?)

The same product (Avery NXR) sells differently to each segment because the deciding factors differ. Our marketing has gotten better as we've learned to segment.

What buyers consistently underestimate

Across segments, buyers tend to underestimate:

1. How much workflow understanding matters.

Buyers often think the platform will figure out their workflows. Reality: the platform is good at executing well-defined workflows. Defining them is still mostly human work.

The customers who succeed are the ones who invest time in workflow definition. The ones who treat the platform as a magic button get disappointed.

2. How much iteration matters.

First-pass agent configurations are rarely great. Iteration over 4-8 cycles produces the working version. Buyers who give up after iteration 2 conclude the platform doesn't work.

Buyers who commit to iteration get value.

3. How much team buy-in matters.

Solo champion adopts. Brings to team. Team either embraces or rejects.

If the team rejects, even great technology doesn't get used. Buyers underestimate the human side of adoption.

4. How much existing process matters.

Agents don't operate in vacuum. They integrate with email, CRM, file systems, communication tools. Buyers underestimate how their existing process affects what's possible.

What buyers consistently overestimate

1. How much capability difference there is between platforms.

Buyers ask "is Avery NXR better than Lindy/Relevance/CrewAI?" The question assumes one is "better." Reality: each is best for specific contexts.

Buyers who pick based on architectural fit (does this match my requirements?) succeed. Buyers who pick based on perceived "best" often pick wrong.

2. How much frontier models matter for their use case.

Buyers default to "I need GPT-4 class reasoning." Reality: most operational workloads don't need it. Local 7-13B models match frontier for what they're actually doing.

Buyers who try local before assuming they need frontier discover this. Buyers who assume frontier pay more for capability they don't use.

3. How much technical complexity is required.

Buyers think AI agent platforms need engineering teams. Reality: most operational AI doesn't.

Buyers who try platforms designed for non-engineers (us, others) discover they can succeed without engineering. Buyers who assume engineering need either hire engineers or don't deploy.

4. How quickly things will be ready.

Buyers think "we'll have agents running in two weeks." Reality: 2-3 months for meaningful adoption across a team.

Buyers who plan for 2-3 months are pleasantly surprised when it's faster. Buyers who plan for 2 weeks panic when month 1 isn't done.

What surprises buyers about Avery NXR specifically

Common "I didn't expect this" feedback:

The Free Desktop tier is genuinely free, not crippled. Some buyers initially distrust the Free tier because they're used to "free tier" meaning "barely functional teaser." When they discover Free Desktop has full functionality (just single-user, on your laptop), they're surprised.

The Excel sync feature is more useful than they realized. We've covered this in [post 167]. Buyers don't expect Excel sync to matter. Then they use it.

The audit ledger pays off in ways they didn't anticipate. Buyers initially see audit as compliance overhead. They discover it as their debugging + investigation tool over time.

Pricing is what it says it is. Buyers expect hidden costs (usage tiers, enterprise upsells). When pricing turns out to be exactly $29/user/month flat, some buyers double-check because it seems too simple.

The visual builder + YAML duality is more important than they thought. Mixed teams (engineers + non-engineers) didn't realize how much friction the single-surface model creates until they had both.

What we hear that we wish we didn't

Some patterns of buyer concerns we hear that suggest broader problems:

"My CEO wants AI deployed but doesn't want to budget for it."

Common. Reveals leadership inconsistency. Hard for us to fix. Leads to deployments that lack support and fail.

"We tried [cloud-LLM platform] last year, it was bad, so we gave up on AI."

The bad experience with one platform created blanket distrust. We have to overcome that distrust, which takes time.

"Our IT department won't let us deploy anything new."

Sometimes legitimate. Sometimes IT department isn't actually objecting, they're just slow to evaluate. We help buyers navigate this when we can.

"My team is afraid AI will replace them."

Sometimes legitimate. Often a sign that leadership hasn't communicated well about how AI fits the org. Hard for us to address directly.

"I want to deploy but I'm not sure how to start."

Common. Reveals lack of decision-making process. We've gotten better at giving frameworks (60-second test, starter pack, etc.) to help buyers start.

What we've changed because of interviews

Specific product/marketing changes that came from customer interviews:

→ The 3-agent starter pack approach [post 171] came from realizing new users were overwhelmed by all 7 templates.

→ Better cost projection content came from interviews where buyers said "I know agents save money but I can't justify it specifically."

→ More comparison content came from buyers saying "I'm comparing you to [competitor], help me understand the difference."

→ The 60-second test [post 196] came from buyers asking "what should I automate first" repeatedly.

→ Better post-purchase onboarding came from interviews with newly-converted customers who felt lost in their first week.

Customer interviews are how we keep the product and content aligned with what buyers actually need.

What buyers say to us privately

A few patterns of feedback we hear privately that don't always make it into formal interviews:

"We picked Avery NXR partly because we like your founder/team." Trust and personality matter more in B2B than companies admit.

"Your honesty in marketing is unusual." When we say what we're NOT good at, buyers notice. Some appreciate it. Some find it confusing (they expect more marketing).

"Your prices are too low — are you going to survive?" Genuine concern from larger buyers. They worry that low pricing means we'll go out of business.

"We're nervous about local AI being good enough." Pre-deployment skepticism about local model quality. Usually resolves after pilot.

"We wish you had X feature." Specific feature requests. We track them all. Sometimes they become roadmap. Sometimes they don't.

The pattern across all 200 interviews

If we had to distill 200 interviews into one observation:

Most buyers underestimate how much THEIR work shapes what they need from an AI agent platform.

They think the question is "what's the best AI platform?" The real question is "what's the platform that best fits MY specific workflows, team, constraints, and goals?"

We've gotten better at helping buyers figure out the second question. The buyers who answer it well end up successful with whichever platform fits — often us, sometimes a competitor.

What we're going to keep doing

Customer interviews are foundational for us. We'll keep doing them, probably another 200+ in the next year.

If you're a customer, prospect, or someone who decided not to buy — we'd love to talk. Honest feedback is more valuable than positive feedback.

The product gets better when we hear honestly what's working and what isn't. The marketing gets better when we hear how people actually decide. The strategy gets better when we hear what people wish existed but doesn't.

→ avery.software — Free Desktop tier. Built by founders who keep listening to customers.