In the last six months, three AI-native CRM startups raised a combined $156M. Monaco closed a $35M Series A led by ex-Founders Fund partner Sam Blond. Lightfield hit a $300M valuation on $81M in total funding. Dreamteam locked down roughly $40M from Accel. Each company claims to be building the CRM that Salesforce should have built ten years ago.
They are not wrong about the problem. Salesforce is a database with a UI that was designed in 2004 and retrofitted with AI features that feel bolted on because they are. HubSpot is friendlier but architecturally similar: a record system with automation layered on top rather than woven through. The new wave of startups sees this gap clearly.
What they get right, where they fall short, and what it means if you are evaluating platforms right now.
Monaco: The Outbound Machine
Monaco is built by Sam Blond, who ran sales at Brex and spent time at Founders Fund backing go-to-market companies. The pitch: an AI agent that handles outbound prospecting end to end. Research the account, write the email, personalize it from public signals, send the sequence, book the meeting. The rep shows up to a qualified conversation instead of spending four hours finding someone to talk to.
Monaco raised $35M because investors believe outbound is the first sales function that AI can fully automate. They are probably right. Outbound prospecting is repetitive, data-intensive, and follows patterns that LLMs handle well. The average SDR sends 150 emails per day with a 2% reply rate. A well-tuned AI agent can match that volume at higher personalization depth, and it never takes a sick day.
Monaco treats outbound as a complete workflow, not a feature. The AI does not just draft an email and wait for approval. It researches, writes, sends, tracks, follows up, and books. That end-to-end approach is correct. The gap: Monaco is outbound only. Once the meeting is booked, you need a separate CRM to manage the deal, a separate tool for call intelligence, a separate tool for forecasting. The data fractures immediately.
The limitation is scope. Monaco does one thing well, but sales teams do not only do one thing. After the meeting is booked, the deal enters a pipeline that Monaco cannot manage. Call recordings go to a separate tool. Forecasts live in another. Commission calculations happen in a spreadsheet. Monaco solves the top of the funnel and ignores everything downstream.
For teams with a healthy mid-funnel and close process, Monaco could be a strong addition. For teams trying to consolidate their stack, it is another point solution.
Lightfield: The Memory Layer
Lightfield raised $81M and is valued at $300M. Their thesis is "complete customer memory." Every interaction, every email, every call, every support ticket, every product usage event gets indexed into a unified timeline. The AI then uses that complete context to surface what matters: this account had a support escalation last week, do not send the upsell sequence yet. This champion changed jobs, their replacement has no context on why they bought.
The concept is sound. Most CRM data is incomplete because most CRM data depends on reps entering it. Reps enter maybe 30% of relevant interactions. The other 70% lives in email threads, Slack messages, call recordings, and support tickets across five different tools. Lightfield's bet is that if you capture everything automatically, the AI can reason across the full picture.
Lightfield ingests data from external tools rather than replacing them. This means you still pay for Gong, Outreach, Zendesk, and ZoomInfo. Lightfield sits on top as a unification layer. At $300M valuation on $81M raised, investors are betting that the memory layer is valuable enough to justify keeping the underlying tools. That is a reasonable bet for enterprises that cannot rip and replace. It is a questionable bet for teams that want to actually reduce their tool count and cost.
The challenge with the "memory layer" approach is that it creates a new dependency without removing old ones. Your stack goes from 11 tools to 12. The data is more accessible, but the cost is higher and the integration surface area increases. When Lightfield's Gong connector breaks, your customer memory has a gap. When Outreach changes their API, someone has to fix the integration.
Lightfield is right that unified context is the key to useful AI in sales. Where they diverge from our view: we think unified context requires a unified platform, not a unification layer on top of fragmented tools.
Dreamteam: The Workflow Engine
Dreamteam closed approximately $40M from Accel. Their angle is workflow automation. Not just sequences (send email, wait, send email), but complex multi-step workflows that span departments. When a deal closes, automatically trigger onboarding in CS, update the forecast, calculate the commission, notify the team, and create the renewal opportunity. All from a single trigger.
This is the right instinct. The biggest time sink in revenue operations is not any single task. It is the handoffs between tasks, between tools, between teams. A deal closes in the CRM, but someone has to manually create the onboarding ticket, someone else has to update the forecast spreadsheet, someone else has to calculate the commission. Dreamteam wants to automate all of that.
Dreamteam's workflows are powerful but they rely on integrations with existing tools. The workflow triggers an action in Salesforce, sends a message through Slack, creates a ticket in Zendesk. If any of those tools change their API or experience downtime, the workflow breaks. Building workflows on top of fragmented tools is better than no automation, but it is still building on sand. The workflows are only as reliable as the weakest integration in the chain.
Accel backed Dreamteam because the RevOps automation market is real. Companies spend 15-25% of a RevOps FTE's time on manual handoffs between tools. Automating those handoffs saves real money. The question is whether you automate across a fragmented stack or consolidate the stack so the handoffs disappear entirely.
What "AI-Native" Actually Requires
Every one of these startups calls itself "AI-native." The term has become so overused that it risks meaning nothing. So here is what AI-native requires architecturally, and where the new wave meets or misses the bar.
First, the AI must operate on a single data model. Not a data lake. Not an aggregation layer. A single schema where contacts, deals, activities, calls, emails, support tickets, and product usage all share foreign keys. When the AI looks at a deal, it should see every related interaction without querying five different APIs. Monaco has this for outbound data. Lightfield tries to simulate it through ingestion. Neither has it for the full revenue lifecycle.
Second, the AI must be able to take action, not just analyze. A system that tells you "this deal is at risk" but cannot do anything about it is a dashboard, not an agent. The AI needs permission-scoped execution: the ability to send an email, update a field, create a task, or trigger a workflow based on what it finds. With audit trails on every action so you can see what happened and reverse it if needed.
Ask any AI CRM vendor: can a sales rep's AI assistant see data from another org on the same instance? If the answer involves "well, in theory, no" or requires a manual configuration step, the platform is not AI-native. Database-enforced multi-tenancy through Row Level Security is not optional. Every query, every AI tool call, every data access must be scoped to the org at the database level. Not the application level. Not the API level. The database level. This is the difference between "we handle security" and "the architecture enforces security."
Third, the AI must be provider-agnostic. Any platform locked to a single LLM provider is making a bet that provider will remain the best option for every use case indefinitely. That is a bad bet. Model performance shifts every quarter. A platform built on GPT-4 in 2024 might need Claude for reasoning tasks and Gemini for multimodal analysis by 2027. Provider lock-in is an architectural debt that compounds.
Fourth, the tool system must be typed. The AI cannot reliably execute actions if the inputs and outputs are untyped strings. Every tool the AI calls needs a schema definition (Zod, JSON Schema, or equivalent) that validates inputs before execution and structures outputs for the next step. Untyped AI tools are the 2026 equivalent of untyped API endpoints: they work in demos and break in production.
Where Revian Fits
We built Revian to be the platform these startups are each building a piece of. 33 capabilities in a single codebase. 400,000+ lines of TypeScript. 119 AI tools, each with Zod-validated schemas. 606 API routes. Database-enforced multi-tenancy through Postgres Row Level Security. Full audit trail across 279 mutation paths. Provider-agnostic AI that can swap models without changing a line of application code.
Monaco's outbound automation? That is our AI SDR capability, one of 33. Lightfield's complete customer memory? That is what happens when CRM, call intelligence, support ticketing, email tracking, and deal rooms all share the same data model. Dreamteam's workflow engine? That is our automation builder, operating on native data instead of cross-tool integrations.
We are honest about where these competitors push us. Monaco's focus on outbound depth is impressive, and it has forced us to think carefully about how our AI SDR handles edge cases in personalization. Lightfield's emphasis on capturing every interaction is right, and the architecture question of AI-native vs. AI-augmented matters more than most buyers realize. Dreamteam's workflow UX is clean. Competition makes everyone better.
Monaco covers roughly 3 of the 33 capabilities a revenue team needs. Lightfield covers about 5, primarily around data unification and intelligence. Dreamteam covers 6-8, focused on automation and workflow. Each does their slice well. But a 50-person sales team needs CRM, sequences, call intelligence, enrichment, scheduling, proposals, forecasting, commission tracking, customer success, deal rooms, support ticketing, and analytics. Getting all of that from one platform at $149/user/month costs less than getting three of them from a funded startup at $200/user/month each.
What This Means for Buyers
If you are evaluating CRM platforms right now, the market is more interesting than it has been in a decade. Salesforce is vulnerable for the first time. HubSpot is fighting on two fronts: defending against Salesforce moving downmarket and against AI-native startups moving up. The new entrants are well-funded, technically capable, and solving real problems.
Three questions to ask any vendor in this space:
How many of the 33 revenue capabilities do you cover natively, without integrations? The answer determines how many other tools you still need. The 2026 revenue platform comparison breaks down the capability coverage across nine vendors.
What is your data model? If the answer is "we integrate with your existing tools," you are buying a layer, not a platform. Layers add cost without removing complexity. A unified data model means the AI can reason across everything without API calls to five different systems.
What happens when the AI makes a mistake? Every AI system will make mistakes. The question is whether you can see what happened (audit trail), undo it (rollback), and prevent it from happening again (permission controls). If the vendor does not have clear answers on all three, their AI is not production-ready.
The AI agent war is real, and the new entrants are building products that are genuinely better than what Salesforce ships today. The question is not whether AI-native will win. It will. The question is whether you want to assemble an AI-native stack from five funded startups or get it from one platform that covers the full revenue lifecycle. That question has a financial answer, and the YC data supports the same conclusion: consolidation wins.
See the full platform
33 capabilities. One codebase. One data model. See how Revian compares to assembling point solutions.
Request Access