Potential growth

Where three fast-growing markets meet

Synapses sits where observability, AI operations and AI-feature monitoring meet. The AI-specific part is growing roughly two to three times faster than classic monitoring.

Research firms define these markets differently and reach very different totals, so we show a range of sources rather than the single most impressive number. All figures were last checked in October 2026.

Market size

Market size and growth rate

Classic observability is a large market growing steadily at about 11–16% a year. AI and LLM observability is smaller but growing at about 23–36% a year.

Classic observabilityAI-specific segments

Market growth per year (CAGR) by segment, with size and forecast
Market	Growth per year	Size and forecast
Observability products^[1]Gartner	11.1% (2021–2028)	Tools to see the health and behaviour of apps and infrastructure. Forecast ~$14.2B by 2028.
Observability platforms^[2]Mordor Intelligence	15.9%	Unified observability platforms. Size $2.9B (2025), Forecast $6.1B by 2030.
AI observability^[4]Meticulous Research	22.8%	Monitoring AI models and agents in production. Size $1.25B (2025), Forecast $13.9B by 2036.
AIOps^[3]The Business Research Company	~30%	Using AI to run IT operations, including automated diagnosis and response. Size $11.1B (2025) → $14.4B (2026), Forecast $41.6B by 2030.
LLM observability platforms^[5]The Business Research Company	36.2%	Monitoring, tracing and evaluating LLM apps. Size $1.97B (2025), Forecast $9.26B by 2030.

It's a bit like the car market: overall it grows slowly, while electric cars inside it grow much faster.

Synapses doesn't compete for the whole observability market on day one. Its wedge is the overlap of LLM observability (36% CAGR) and AIOps-style automated root-cause analysis, where incumbent tools are newest and least settled.
Totals differ because some firms count only software licences for a narrow product type, while others include services and adjacent categories. One older report projects AIOps at around $645B by 2030 using a much broader definition [6]. That's an outlier, so we don't lead with it.
The honest framing: a multi-billion-dollar market, with the AI segment growing at 20–36% a year.

Audience

Who it's for

Start with small and mid-sized software teams that ship AI features, then grow into platform teams and larger companies.

AI-native startups and scale-ups
5–200 engineers building products with LLM features, often using AI coding tools
Pain: AI costs they can't explain, wrong AI answers, incidents with no on-call team, AI-written code shipping fast
Why Synapses fits: One SDK covers infra, users, AI cost and security; the wrapper is one line; automated diagnosis replaces a missing SRE team
First
Platform and SRE teams at mid-sized companies
Teams that run production for a whole company
Pain: Too many tools, alert fatigue, slow root-cause analysis
Why Synapses fits: One causal graph across signals; findings that lead to an action
Second
Engineering leaders and founders
CTOs, VPs of Engineering
Pain: Can't connect outages and AI spend to revenue
Why Synapses fits: Impact shown in users and money, not just latency
Buyer persona
Large enterprises
1,000+ employees, regulated industries
Pain: Compliance, data residency, change control
Why Synapses fits: Policy-bound autonomy and audit trails (needs a self-hosted option and certifications first)
Later

84%

of developers use or plan to use AI tools; about half of professional developers use them every day.^[7]

75.8%

of developers don't plan to use AI for deployment and monitoring. AI speeds up writing code, not running it. That gap is where Synapses fits.^[7]

Engineers use Synapses every day. Engineering leaders decide to adopt it, because it shows incidents and AI spend in terms of money.

Opportunities

Main growth opportunities

AI cost and quality monitoring is a new budget line.
Every company adding LLM features now has a monthly AI bill and a quality risk it can't easily see. Synapses shows cost per feature, per user and per model, and can cap spend automatically.
Why it's real: LLM observability is the fastest-growing segment above (36.2% CAGR). Gartner forecasts that half of generative-AI deployments will include LLM observability investment by 2028, up from 15%.^{[5, 8]}
Example: Like an itemised phone bill for your AI: you see which feature costs what, and calls stop when the limit is reached.
Teams want fewer tools.
Replacing several separate tools with one platform is an easier sale than adding yet another tool.
Why it's real: In a New Relic survey of 1,700 teams, the average number of observability tools fell from 6 in 2023 to 4.4 in 2025. Only 27% of organisations had full-stack visibility, and 52% plan to consolidate onto a unified platform within two years. Gartner has published guidance on cutting observability tool sprawl.^{[9, 10]}
Faster root cause saves real money.
Most of an outage is spent working out what went wrong, not fixing it. Automated diagnosis shortens that part.
Why it's real: Over 90% of mid-sized and large firms say one hour of downtime costs them more than $300,000. New Relic's 2025 report puts the median cost of a high-impact outage at about $2M per hour.^{[11, 12]}
Example: If a team cuts diagnosis from 45 minutes to 5 on a $300,000-an-hour outage, it saves roughly $200,000 on that one incident.
Security for AI-written code.
AI coding tools are fast but often leave out basic protections. Synapses watches the running app for what's missing, such as authorisation checks.
Why it's real: Veracode tested over 100 LLMs on 80 coding tasks. In 45% of cases the generated code contained a security weakness.^[13]
Term: IDOR (insecure direct object reference)
A flaw where changing an ID in a request lets one user see another user's data. Example: changing /invoice/1001 to /invoice/1002 and seeing someone else's invoice.
Extending from production back to code.Roadmap
Linking production problems to the commit, CI run and deployment that caused them opens a second market: CI and deployment intelligence.
Why it's real: Developers increasingly ship code they didn't write line by line, so "which change broke this?" is getting harder to answer by hand.
Term: CI (continuous integration)
The automatic build-and-test run that happens on every code change.
Plugging into coding agents through MCP.
Giving AI coding assistants live production context through MCP puts Synapses inside tools developers already use every day.
Why it's real: AI-enabled editors are already in wide use; for example, 18% of Stack Overflow respondents use Cursor.^[14]
Term: MCP (Model Context Protocol)
A standard way for AI assistants to query other systems, like a universal plug.

Trends

Trends that support these opportunities

More AI in products, more AI-written code, more pressure on cost, and open standards all push in Synapses' direction.

TrendWhat it meansEvidence

AI moving from experiments into production: What it means: Monitoring AI becomes a requirement, not a nice-to-have; Evidence: AI observability is growing as companies move AI into production^[4]
AI-assisted coding is mainstream: What it means: More code ships faster, with more hidden flaws; Evidence: ~51% of professional developers use AI tools daily; 45% of AI-generated code samples were insecure in testing^{[7, 13]}
Tool consolidation: What it means: Buyers prefer one platform over many point tools; Evidence: 52% plan to consolidate within two years^[9]
Agentic AI in software: What it means: Software increasingly acts on its own, so it needs guardrails and audit trails; Evidence: Gartner expects 33% of enterprise apps to include agentic AI by 2028, up from under 1% in 2024^[15]
OpenTelemetry as the standard: What it means: Teams can switch vendors without re-instrumenting, which lowers the barrier for a newcomer; Evidence: OTel standardisation of AI telemetry is cited as a growth opportunity^[4]
AI observability merging with AI security: What it means: One product for "is it working?" and "is it safe?"; Evidence: Market analysts cite this convergence as an opportunity^[4]

Risks

Key risks

The biggest risks are trust in autonomous actions, strong incumbents, and the cost and privacy of storing so much data.

RiskWhat could go wrongHow Synapses reduces it

Trust in autonomy: What could go wrong: Teams won't let an AI change production; one bad automatic action destroys trust; How Synapses reduces it: Start in read-only or PR-only mode; confidence scores; an evidence packet for every action; kill switch
Agentic AI hype backlash: What could go wrong: Buyers grow sceptical of "AI agent" claims; How Synapses reduces it: Show real evidence and measured outcomes, not slogans. Gartner expects over 40% of agentic AI projects to be cancelled by end of 2027 because of cost, unclear value or weak risk controls, and says only about 130 vendors offer genuine agentic capabilities^[15]
Incumbent competition: What could go wrong: Large vendors (Datadog, New Relic, Dynatrace, Grafana) are adding LLM monitoring and AI features; How Synapses reduces it: Compete on the cross-domain causal graph and on taking action, not on single features
Data volume and cost: What could go wrong: Storing traces, replays and prompts gets expensive; How Synapses reduces it: Sampling, retention controls and usage-based limits. Analysts name telemetry volume and cost as a market restraint^[4]
Privacy: What could go wrong: Prompts and session replays can contain personal data; How Synapses reduces it: Mask data by default before it reaches the graph; customer-controlled retention. Privacy of logged prompts is named as a market restraint^[4]
Measuring AI answer quality: What could go wrong: Without a known correct answer, quality scores can be wrong; How Synapses reduces it: Present scores as estimates with confidence; let teams add their own checks. Analysts call this a major challenge^[4]
Developer trust in AI output: What could go wrong: Engineers double-check AI results and may ignore them; How Synapses reduces it: Always show the evidence behind a finding. In 2025, more developers distrusted AI accuracy (46%) than trusted it (33%)^[7]
Integration breadth: What could go wrong: The product is only as good as the data it can connect; How Synapses reduces it: Lean on OpenTelemetry; focus on a few deep integrations first (GitHub, GitHub Actions, major LLM providers)

Comparison

What Synapses does differently

Synapses

Compared with Datadog: Synapses connects LLM cost and quality to infrastructure, users, security and revenue in one causal graph.
Compared with LangSmith: LangSmith centres on LLM and agent development and evaluation; Synapses adds infrastructure, user and business context, and acts on it.
Compared with Snyk: Synapses focuses on runtime behaviour and execution-path context for AI-generated apps.
Compared with Helicone: Helicone is gateway-centric; Synapses intercepts in-process and adds full-stack context and remediation.
Compared with PagerDuty: PagerDuty: page → investigate → fix. Synapses: detect → trace the cause → propose or act → verify.

Sources

All figures were last checked in October 2026.