Three customers. Three different deployments. Same product.

One of the harder things about positioning Avery NXR is that "who uses it" looks different across customer segments.

A solo consultant uses it differently than a 50-person team. A 50-person team uses it differently than a 500-person regulated enterprise.

Same product. Three very different deployments. Here's what each looks like.

Customer 1: Solo consultant

Profile: Senior product consultant. Works with 8-12 clients at a time. Works from a MacBook Pro M3 with 36GB RAM. No engineering team, no IT, no procurement process.

Tier: Free Desktop. $0/month.

Deployment: → Ollama running locally → Avery NXR desktop app → Model: Qwen2.5-Coder 7B Q4 → 6 agents configured

Agents: → Anna template — daily AI/product industry digest → Sophia template — meeting follow-up after client calls → Custom proposal drafter → Custom client status report → Yuki template — competitor monitoring (the clients' competitors) → Custom invoice processor

Connector usage: → Gmail (OAuth) → Google Calendar (OAuth) → Notion (OAuth, project notes) → QuickBooks (API key) → Various API-key for niche tools

Value delivered: → ~10 hours/week of operational work absorbed by agents → Effective consulting capacity expanded from 8 to 12 clients → Annual cost: $0 → Equivalent to hiring a part-time admin at $30K/year

Why this works: The solo consultant doesn't have engineering bandwidth to build agent infrastructure. Free Desktop tier means no negotiation, no contract, no procurement. Install, configure, run. Local AI means no cost concerns for personal use. Excel sync means client data spreadsheets work as data sources.

Customer 2: 50-person mid-market SaaS company

Profile: B2B SaaS company. 50 employees. Series A funded. Has small engineering team but no dedicated AI infrastructure team. Cost-conscious about subscription bloat.

Tier: Avery NXR Pro. $29/user/month × 35 users = $1,015/month. About $12K/year.

Deployment: → Pro tier with cloud deploy → Deployment target: Railway (their existing infrastructure) → Avery NXR runtime on shared cloud instance → Model: Qwen2.5-Coder 7B + occasional Consult Mode escalation to Claude (via BYOK) → 18 agents deployed across departments

Agents (by team):

Engineering: → Server health (Liam template + custom monitors) → PR status digest → Incident response triage

Sales (8 reps): → Pipeline digest (Carlos template) — personalized per rep → Inbound lead qualification → Quote drafting (with Excel pricing sync) → Account research before discovery calls

Marketing: → Competitor monitoring (Yuki template) → Content idea generation → Newsletter drafting

Customer support: → Ticket triage (Priya template) → Knowledge base maintenance

People ops: → Resume screening (Marcus template) → Onboarding email sequences → Interview scheduling

Operations: → Invoice processing → Vendor management

Connector usage: → HubSpot (OAuth, CRM) → Linear (OAuth, engineering) → Slack (OAuth, multiple channels) → Gmail (OAuth) → Zendesk (API key) → ~12 other connectors

Value delivered: → Replaced ~$25K/year in AI-flavored SaaS subscriptions → Multiple workflows that didn't have SaaS coverage (faster than building from scratch) → Audit ledger for SOC 2 compliance questions → Data residency answer for enterprise prospects asking

Why this works: Mid-market needs centralized deployment (not laptops) so agents run continuously regardless of who's at their desk. Pro tier provides cloud deploy. Audit ledger handles compliance. Cost is flat $12K/year vs. competitor cloud-LLM platforms that would project $40-60K for the same workload.

Customer 3: 500-person regulated enterprise

Profile: Financial services firm. 500 employees. Compliance-heavy industry. Strict data residency requirements. Existing security review process for any new vendor.

Tier: Avery NXR Enterprise. Custom pricing.

Deployment: → On-premises deployment via SSH on dedicated infrastructure → Air-gapped from the public internet for sensitive workloads → Local model running on internal GPU server (Qwen2.5-Coder 14B + DeepSeek-R1-Distill 8B for reasoning tasks) → NO Consult Mode (compliance doesn't allow opt-in cloud escalation for sensitive data) → 40+ agents across business units

Agents:

Compliance: → Customer communication review (flag potential compliance issues) → Trade documentation review → Regulatory filing assist

Operations: → Document classification at intake → Reconciliation drafting → Vendor onboarding processing

Customer service: → Ticket triage with strict routing rules → Account inquiry first-pass drafting

Risk: → Trade alert investigation → Audit prep documentation

Connector usage: → Internal databases (Postgres, custom data warehouse) → Internal email (Microsoft Exchange) → Internal collab (Teams + SharePoint) → Custom HTTP connectors to internal systems

Value delivered: → Operational work absorbed across 40+ workflows → Audit ledger satisfies internal compliance requirements → Data never leaves their infrastructure (compliance answer is clean) → No vendor LLM API calls = no third-party data exposure → Estimated 4-6 FTE worth of operational labor automated

Why this works: Regulated enterprise needs air-gapped deployment + structured audit + no external API calls. Enterprise tier provides all of it. On-prem deployment fits their existing infrastructure pattern. Local models eliminate the cloud LLM data exposure problem entirely. Pricing is custom (substantial) but cost-per-FTE-equivalent is excellent.

What's the same across all three

→ Same agent runtime → Same template library → Same connector library → Same audit ledger architecture → Same configuration model (visual builder + YAML) → Same model integration pattern (Ollama for local, BYOK for cloud)

The product is the SAME. The deployment characteristics flex to match the customer's environment.

What changes by tier

Free Desktop: → Runs on user's laptop → All connectors → All 7 templates → Local AI via Ollama → Audit ledger → Single user → $0/month

Pro: → Cloud deploy (Vercel / Railway / SSH) → Multi-user collaboration → Premium models + Consult Mode (BYOK) → Team-level admin → $29/user/month

Enterprise: → On-premises capable → Air-gap capable → Custom SSO integration → Compliance reporting → Dedicated support → Custom pricing

The flex from "free for solo on a laptop" to "enterprise air-gapped deployment" is unusual in agent platforms. Most products pick one of those ends and optimize there.

The thesis behind one product / three deployments

We believe local-first is the right architecture for operational AI across customer sizes. The deployment context changes (laptop vs cloud vs on-prem), but the architectural principle holds.

A solo consultant benefits from local-first because cost stays at $0. A mid-market company benefits because cost stays predictable and data residency works. A regulated enterprise benefits because air-gap is structurally possible.

If we'd built for one segment, the other two wouldn't have a fit. Because we built for the local-first principle, the same product flexes across the segments.

What this means for evaluating fit

If you're considering Avery NXR, the question isn't "is this product for companies like mine?"

The question is "does local-first AI architecture match how my company wants to do AI?"

If yes — there's a tier that fits.

If no — pick a cloud-first platform optimized for the segment you're in.

→ avery.software — Free Desktop, Pro, Enterprise. Pick the tier that matches your context.