I've been having the same conversation with association executives for two years now. The current AMS is frustrating. The vendor's roadmap is vapor. The board is asking about AI. The membership team is exporting to spreadsheets to do anything analytical. And the conclusion the leadership team always arrives at is "we need a new AMS." It’s an understandable conclusion.

Three numbers from the latest industry research finally explain why this keeps ending badly.

In Marketing General Incorporated's 2025 Membership Marketing Benchmarking Report, only 11 percent of respondents said their association offers a very compelling value proposition, down from 13 percent the two years before, and 45 percent reported a membership increase over the past year. In MGI's 2026 report, that share fell to 38 percent.1 Virtuous's 2026 Nonprofit AI Adoption Report (a benchmark study of 346 organizations) found that 92 percent of nonprofits are using AI but only 7 percent report major strategic impact.2 And Naylor's 2025 Association Benchmarking Report found the share of associations using no AI fell from 64 percent to 42 percent in a year.3

These numbers are the same story. Associations are racing to adopt AI, racing to grow membership, racing to deliver better experiences, and quietly discovering that the platforms they sit on can't actually deliver any of it. The reflexive response has been "buy a more modern AMS." The honest response is to ask whether buying any AMS is the right move at all.

The math on the monolithic AMS no longer works

For thirty years, association technology has been organized around one assumption: the AMS is the source of truth. One vendor, one platform, one database, owning members, dues, events, communications, community, learning, certifications, and reporting. The pitch was integration. Buy everything from us and it will all work together.

The pitch has aged badly. Members expect Netflix-grade personalization, one-click renewals, and mobile-native experiences. Boards expect AI-driven insights and real-time decision support. Staff expect modern tooling that doesn't require an IT ticket to pull a list. None of this is what monolithic AMS platforms were built for. Bolting it on after the fact has produced exactly the results you'd expect: half the features work, none of them work well, and the platform fights its own users at every turn.

Here's what frustrates me about the standard response. The strategic problem is structural, not implementation-related. A monolithic AMS is a data prison. Member data lives inside the platform's proprietary schema, accessible only through the vendor's tooling, integrated with other systems through whatever connectors the vendor decided to prioritize. You cannot run modern AI on this architecture. You cannot deliver a unified member experience across systems. You cannot rapidly adapt as new tools emerge. The vendor's roadmap is the ceiling on your strategy.

This isn't hypothetical. A smaller MGI mid-year survey in July 2025 found 36 percent of associations reporting a membership decline, up from 26 percent in January 2025.4 The associations losing members are disproportionately the ones who can't personalize engagement, can't identify at-risk members early, and can't orchestrate journeys across the systems members actually use.

What's actually replacing the AMS

The pattern I keep seeing at the most progressive associations isn't "buy a different AMS." It's "stop thinking in AMS terms entirely." The right mental model is the modern e-commerce stack, which abandoned monolithic platforms a decade ago in favor of composable architecture: a backbone CRM, dedicated payment infrastructure, best-of-breed point solutions for specialized functions, and a data layer tying it all together. Gartner predicted in 2020 that composable commerce adopters would outpace competitors by 80 percent on the speed of feature implementation by 2023.5 That horizon came and went without anyone auditing the claim, and the number now circulates with a 2026 date bolted on, so read it as the sector's statement of intent rather than a measured result. The direction is the part that held: e-commerce rebuilt itself around composability and did not go back.

The composable member stack has six core layers, each chosen for its strength rather than its membership in a vendor suite.

flowchart TB
    subgraph Old["MONOLITHIC AMS (legacy model)"]
        AMS["Single AMS Platform
Members + Dues + Events + Email + Community + LMS
(everything in one proprietary database)"]
    end
    subgraph New["COMPOSABLE MEMBER STACK (modern model)"]
        CRM["CRM Backbone
(Salesforce / HubSpot)"]
        Pay["Payment Infrastructure
(Stripe / Adyen)"]
        Comm["Communications
(Klaviyo / Customer.io)"]
        Comty["Community
(Higher Logic / Discourse)"]
        Learn["Learning
(Thought Industries / Path LMS)"]
        Events["Events
(Cvent / Swoogo)"]
        Data["Data Layer
(Snowflake / BigQuery + dbt)"]
        AI["AI and Activation
(LLM APIs + Reverse ETL)"]
        Data --- CRM
        Data --- Pay
        Data --- Comm
        Data --- Comty
        Data --- Learn
        Data --- Events
        AI --- Data
    end
    Old -.-> New

The CRM backbone. Salesforce dominates here; HubSpot is increasingly viable for smaller associations. The CRM holds the canonical member record, contact relationships, and core engagement history. Salesforce says 96 percent of its nonprofit customers achieve positive ROI, citing its own 2026 Customer Success Metrics.6 Salesforce's 2025 launch of Agentforce Nonprofit (formerly Nonprofit Cloud) embeds AI agents directly into core workflows, which no traditional AMS has matched.7

Payment infrastructure. Stripe, Adyen, or specialized association-friendly options. Better rates, better fraud handling, better global support, and dramatically better data on transactions than any AMS-bundled payment processor I've ever audited.

Communications. Email and SMS through modern marketing automation (Klaviyo, Customer.io, HubSpot, Marketing Cloud) instead of the AMS email module. The capability gap between specialized communications platforms and AMS-bundled email is now wider than at any point in the last decade. It's embarrassing for the AMS vendors, honestly. They just can't keep up.

Community. Higher Logic remains the leader. Discourse and modern alternatives are rising fast. The community lives independently and feeds engagement signal back to the data layer.

Learning and events. Specialized platforms (Thought Industries, Path LMS, Cvent, Swoogo) for the functions that genuinely benefit from specialized capability.

The data layer. This is the piece most associations underestimate, and the piece that determines whether the rest of the stack actually works as a system. A modern data warehouse (Snowflake, BigQuery, or Databricks), a transformation layer (dbt), and reverse ETL (Hightouch or Census) connect everything. The warehouse is the system of record for analytics, AI, and cross-system intelligence. I've written about the architecture in detail in Why Your Data Warehouse Is the New System of Record.

What the composable stack enables that an AMS cannot

This isn't a philosophical argument. It's an operational one, and the operational gap widens every quarter.

AI that actually works. Generative AI grounded in real-time member data. Agentic workflows that span systems. Predictive models that use signal from every touchpoint. None of this is achievable inside an AMS data prison. The Virtuous research points the same way on one count: the nonprofits pulling ahead are not simply the ones using AI more, but the ones with basic readiness in place, which it names as governance, measurement, and systematic use.2

Real personalization. Members get experiences shaped by their full activity (what they've read, attended, posted, completed, donated to) instead of the narrow slice their AMS happens to know about. This is where a composable stack's shared data layer shows up most visibly.

Speed to capability. A composable stack lets a team swap or add one component without rebuilding the rest. Translation: when the membership team wants to launch a new program, the technology stops being the bottleneck.

Vendor independence. Every component is replaceable. The CRM backbone is the only sticky decision. Everything else can be swapped without rebuilding the foundation. This is the opposite of the AMS lock-in model that has trapped associations on aging platforms for decades.

Future-readiness for agentic AI. The dominant 2026 conversation in enterprise tech is agentic AI: systems where AI agents act on behalf of users and across systems. Salesforce has rebuilt its nonprofit product around AI agents.7 Associations on composable stacks can start experimenting with agentic AI now, because their data is already reachable outside a single vendor's platform. Associations on monolithic AMS platforms have to wait for their vendor's roadmap. Expect that difference to matter more as agentic tools mature.

The "but we already have an AMS" question

Most associations reading this aren't starting from scratch. They have an iMIS, an Aptify, a Personify, a MemberSuite, or a Salesforce-based AMS already. The composable approach isn't "rip and replace." It's "decompose and modernize."

The pattern that works:

  1. Stand up the data layer first (Snowflake or BigQuery), and load member, transaction, engagement, and demographic data from every system into it.
  2. Build the unified member record (Member 360) in the warehouse, not in any operational system.
  3. Replace the most painful component first. For most associations this is communications, because the AMS email module is almost always the worst part of the AMS.
  4. Add modern community, learning, or events capabilities as best-of-breed alternatives, fed by warehouse data.
  5. Reduce the AMS to its core function (records and dues) while everything else moves to specialized tools.
  6. Eventually, replace the AMS itself with a CRM-backbone architecture, or keep it as a financial system of record while the CRM owns relationships.

This phased approach spreads the work out, and no published benchmark shows whether it costs less overall than a full AMS replacement. It depends on scope. Critically, every phase delivers operational value on its own. There's no "we'll see results in three years" gamble. I've seen too many three-year migrations end up exactly nowhere. This isn't that.

For a deeper look at what this transition actually involves, see my migration playbook.

What this costs (and what it costs not to do it)

What a composable member stack typically includes, by association size. Price each line against your own quotes: there is no published benchmark for the total.

No published study compares the five-year cost of a composable stack with a monolithic AMS replacement, so do not trust anyone who quotes you a percentage saving, including me. The difference that matters is capability: the composable stack improves over time as components evolve, while the monolithic AMS depends on one vendor's roadmap keeping up.

The cost of not doing this is harder to put on a single line. It's the membership decline that comes from undifferentiated experiences. It's the AI capability gap that widens every quarter. It's the staff turnover from working on broken tools. It's the strategic options that quietly close as competitors move ahead. None of these show up in a budget. All of them show up in five years.

What to do next

If you're at the start of an AMS evaluation, the most useful question to ask before issuing the RFP is whether the AMS conversation is even the right conversation. For most associations in 2026, it isn't. The right conversation is about the member stack.

If you're mid-implementation on a traditional AMS replacement, it's worth pausing. Specifically, it's worth asking whether the platform you're moving to is genuinely modern composable architecture or just a more recent monolith. Salesforce-based AMS platforms are the closest thing to composable in the traditional category, and they remain real options. Most other AMS replacements are not.

If you're stuck on a legacy AMS and the next replacement cycle feels overwhelming, the composable approach offers something the AMS replacement playbook does not: a path to modernize without a one-shot bet-the-organization migration. The first phase delivers value in 3 to 4 months. Every subsequent phase delivers value too.

The associations that will thrive over the next decade are the ones whose architecture matches their ambition. The AMS-as-monolith model doesn't. The composable member stack does. Which side of that line your association lands on is a decision being made right now, whether it feels like one or not.

ARYS Intelligence is my practitioner-led technology practice for associations, nonprofits and foundations. I help organizations design composable member stacks that support modern member experience, AI capability, and operational agility. If you’d like to talk through an architecture assessment in confidence, get in touch.