A three-year AI roadmap for an advisor practice is not a vendor selection exercise dressed up in slides. It is an ordered investment thesis — Year 1 productivity, Year 2 integration, Year 3 differentiation — calibrated to the practice archetype (Solo, Ensemble, Multi-Custodian, Wirehouse, OSJ from L4 Ch1 L1), grounded in the readiness audit (L4 Ch1 L2), and budget-anchored to per-advisor-seat economics that survive a CFO's stress test. This lesson installs the roadmap template, the budget anchors per archetype per year, the Marketing Rule and ADV implications of each tier, and the three failure modes (premature Year 3 ambition, Year 1 tool sprawl, Year 2 integration debt) that derail most attempts. By the end you can draft the practice's three-year roadmap on one page, defend the budget to a CFO, and disclose the trajectory to a buyer or an examiner.

Why a Three-Year Horizon — Not Five, Not One

Five-year AI roadmaps for advisor practices are fiction. The technology, the regulators (SEC Marketing Rule 206(4)-1 staff FAQs, FINRA Annual Regulatory Oversight Reports, Reg S-P amendments, NAIC AI Model Bulletin updates), the vendor landscape, and the M&A environment all move faster than a 60-month plan can accommodate. The Schwab 2026 RIA Benchmarking Study notes AI adoption more than doubled between 2023 and 2026; nothing in that trajectory suggests five-year planning is durable. One-year plans, conversely, are reactive: by the time the practice finishes integrating the year-one tools, the deadline has passed without the integration phase that produces real ROI.

Three years is the only horizon that fits the reality of advisor AI in 2026. Year 1 is the productivity year — meeting AI, planning AI, tax extraction — where the practice books concrete time savings (Schwab 2026 data and Zocks' 10+ hours/week claim are the named anchors). Year 2 is the integration year — workflow automation across the CRM / planning / archive / LLM stack, RAG to the firm's vault, Marketing Rule pre-use review queue at scale. Year 3 is the differentiation year — proprietary workflows, agentic tools under FINRA Rule 3110 reasonable-design supervision, AI-native service tiers that change the practice's economics and ADV fee schedule. The three years map cleanly to the L4 chapter sequence: Ch1-Ch2 sets up Year 1, Ch3-Ch4 fortifies through Year 2, Ch5-Ch8 unlocks Year 3.

The roadmap is a living document. It is re-baselined annually after the readiness audit, updated mid-year on material events (new advisor, new custodian, M&A, examination, breach), and filed under the WSP as a compliance artifact under SEC Rule 204-2 and FINRA Rule 4511. The roadmap is the document an SEC examiner will ask for when the question is "what AI process did you follow?" — and the document a sophisticated M&A buyer will request in the L4 Ch8 L2 diligence pack as evidence of AI maturity.

Year 1 — Productivity

Year 1's mission is single-purpose: book defensible time savings against documented baseline workflows. Nothing fancy. Three investment categories, each with a single-vendor-per-category discipline.

Meeting AI deployment. Jump or Zocks (pick one) plus the Smarsh / Global Relay archive integration (L3 Ch10 L1 pipeline). Solo and small ensemble: $50-70/advisor seat/month for Jump or Zocks Standard. Mid-ensemble: $70-100/seat for Jump Enterprise or Zocks Business. Multi-custodian: $100-150/seat for the integration-rich tier. Wirehouse FA: home-office allocation, no direct cost. OSJ: archive-integration tier $200-400/supervised advisor seat for supervisory-grade capture. The deliverable is the L3 Ch10 archive pipeline operational and the L2 Ch2-Ch3 prep/notes/follow-up workflow live across the team.

Planning AI deployment. Holistiplan for tax extraction ($150-300/advisor/month depending on volume), FP Alpha or Wealth.com for estate extraction ($100-300/advisor/month), and the existing planning software (RightCapital, eMoney, MoneyGuidePro) without switching. The deliverable is the L2 Ch4 workflows live: 1040 to tax memo with QCD election workflow, NUA decision for separating executives, estate gap memo from trust/will extraction, retirement income drawdown memo. Cross-references: the L3 Ch2 Roth conversion screen and L3 Ch5 estate audit become possible only after the planning AI is operational.

Enterprise LLM seat licensing. Microsoft Copilot, OpenAI Enterprise, Google Gemini Enterprise, or Anthropic Claude for Work (pick one per the L4 Ch1 L1 archetype-driver framing). Solo: $30-60/seat for the entry enterprise tier. Ensemble: $60-90/seat for the team tier with SSO and SCIM. Multi-custodian: $90-150/seat for the high-volume tier. The deliverable is the team using the LLM under a documented system prompt (L2 Ch8 L1) with the Cardinal Rule verification protocol (L1 Ch2.3) habituated.

Year 1 budget anchors per advisor seat per month, all-in for the AI layer: Solo RIA $400-700, Small Ensemble (3-10 advisors) $600-900, Mid Ensemble (10-30 advisors) $800-1,200, Multi-Custodian $1,000-1,500, OSJ supervisory layer $500-1,000/supervised advisor incremental to RIA-side spend, Wirehouse FA $0 (allocated by home office).

Marketing Rule and ADV implications of Year 1. Year 1 introduces AI tools that touch client NPI and that produce client-facing output. The ADV Part 2A AI disclosure becomes mandatory under the L5 Ch7 framework. The Marketing Rule 206(4)-1 pre-use review process must extend to AI-drafted content (L4 Ch3 L2). The substantiation file for any AI capability claim in marketing materials must be live. The L4 Ch7 L1 AI-washing audit becomes annual. The Reg S-P 17 CFR Part 248 vendor oversight obligation (May 2024 amendments) requires DD for each tool added — the L4 Ch2 L2 40-question questionnaire is the deliverable, retained for each vendor under SEC Rule 204-2 and FINRA Rule 4511.

Year 2 — Integration

Year 2's mission is to turn the Year 1 tools into a system — workflow automation, archive coverage, RAG to the firm's vault, supervisory tooling, Marketing Rule pre-use review at scale. The economics inflect upward (tools-per-advisor count increases) but the ROI compounds. This is the year the L3 capstone deliverable (the practice playbook with 10 named workflows) becomes operational.

Workflow automation. Connect Jump / Zocks to Wealthbox / Redtail / Salesforce FSC / Practifi (CRM ingestion of structured AI output per L2 Ch8 L2). Connect Holistiplan extracted facts to RightCapital / eMoney / MoneyGuidePro. Connect Orion Eclipse to the planning software for portfolio-to-plan sync. Connect Catchlight (or equivalent) leads to the CRM and the marketing automation. Integration spend Year 2: $200-400/advisor/month incremental, paid largely to integration platforms (Beacon, Hubly, custom) and CRM customization labor.

Archive coverage and Rule 4511 retention. Extend Smarsh / Global Relay capture to every channel — Zoom / Teams / Webex recordings, Jump / Zocks meeting summaries with transcripts and advisor edits and signoff trail, LLM prompts and outputs (selective retention per the L3 Ch10 L2 prompt-retention policy), client-portal messaging, and any new agentic-AI artifacts. The deliverable is a documented capture matrix covering every artifact under FINRA Rule 4511 and SEC Rule 204-2. Archive spend Year 2: $30-100/advisor/month incremental.

RAG to firm vault. Connect the enterprise LLM (Copilot / OpenAI Enterprise / Gemini Enterprise / Claude for Work) to the firm's IPS template library, compliance-approved disclosure language, prior Reg BI memo examples, and the firm's "house voice" templates. The L3 Ch9 L2 retrieval-augmented-workflows lesson is the implementation. The deliverable is the model speaking the firm's voice and reusing compliance-blessed language by default. RAG spend Year 2: $50-150/advisor/month for the platform tier plus vector-database costs.

Supervisory tooling. For OSJs and CCO-staffed RIAs, deploy Marketing Rule pre-use review queue with AI first-pass screening (L4 Ch3 L2), Reg BI memo audit tooling, and Rule 4511 supervisory log capturing the entire human-and-AI chain. Supervisory spend Year 2: $300-700/supervised advisor/month incremental for the OSJ layer; $100-300/advisor/month for the RIA-side CCO tooling.

Cyber upgrades. The May 2024 Reg S-P amendments are fully effective by Dec 2025 for large advisers ($1.5B+ AUM threshold) and June 2026 for smaller advisers — Year 2 is when the IRP testing, vendor oversight cadence, and 30-day breach-notification machinery must be operationally live (L4 Ch4 L1). NY DFS 23 NYCRR 500 third-party-service-provider expectations apply to any NY-licensed business. Cyber spend Year 2: $100-300/advisor/month for enterprise IAM (Okta, Microsoft Entra), MDR / SOC-as-a-service, and vendor management tooling.

Year 2 budget anchors per advisor seat per month, all-in: Solo $700-1,100, Small Ensemble $1,000-1,500, Mid Ensemble $1,400-2,000, Multi-Custodian $1,800-2,800, OSJ adds $800-1,500/supervised advisor incremental, Wirehouse FA still $0 direct.

Marketing Rule and ADV implications of Year 2. The RAG to firm vault changes how AI produces client-facing content — every output now references compliance-blessed language, which materially changes the disclosure posture. ADV Part 2A is updated to reflect the deeper AI use across workflows (L5 Ch7 off-cycle amendment if material). The Marketing Rule audit (L4 Ch7 L1) becomes more substantive because the firm now has documented AI capability claims worth examining. The agentic-AI WSP (L4 Ch3 L3) becomes the design target as Year 3 capabilities (action-taking AI) approach.

Year 3 — Differentiation

Year 3's mission is to convert AI maturity into a competitive moat that survives the next wave of RIA consolidation. Proprietary workflows that vendors don't ship. Agentic tools deployed under defensible Rule 3110 reasonable-design supervision. AI-native service tiers that change the practice's economics, the ADV fee schedule, and the conflict-disclosure narrative.

Proprietary workflows. The L5 Ch2 L1 lesson is the framing: where there is no Jump or Zocks yet — legacy planning for blended families, business-owner exit workflows (with the QSBS Section 1202 dual-regime analysis post-OBBBA), multi-generational philanthropic stewardship, expat tax + estate, complex annuity laddering — the firm builds its own prompt library, RAG pipeline, and internal AI assistant. The practice's specialized knowledge (e.g., equity-comp executive niche, family-office trustee-advisor work) gets coded into AI workflows that no shelf vendor matches. Year 3 proprietary-workflow spend: $200-500/advisor/month plus 1-3 FTE-equivalent of internal prompt librarian / AI operations time.

Agentic AI under Rule 3110 reasonable-design. The FINRA 2026 Annual Regulatory Oversight Report devoted a section to agentic AI under Rule 3110. The L4 Ch3 L3 lesson is the WSP framing. Year 3 deployments — AI that places trades (Orion Eclipse rebalance), sends emails on a documented protocol, files certain forms (with human review queues), processes RMDs, initiates ACATs — require pre-trade compliance, post-action review with sampling, kill-switch design, Rule 4511 supervisory log, and the Reg BI documentation chain. Done well, agentic AI is a step-change productivity unlock; done badly, it produces the next FINRA enforcement headline. Agentic AI spend Year 3: $300-700/advisor/month for the platform and supervisory layer.

AI-native service tiers. Year 3 introduces service tiers that didn't exist before AI compressed cost-to-serve. A subscription planning tier for the $200k-$1M household segment, accessible AI-augmented financial planning at lower fee than traditional AUM, with a defined human-review cadence. A premium UHNW tier with agentic AI doing the operational work and the senior advisor freed for judgment. Each tier requires an ADV fee schedule update, an engagement letter rewrite, a conflict-disclosure update (L4 Ch7 L2), and a Marketing Rule audit on any new capability claim. Service-tier design spend Year 3: $150-400/advisor/month for the platform infrastructure.

Year 3 budget anchors per advisor seat per month, all-in: Solo $1,000-1,500, Small Ensemble $1,500-2,200, Mid Ensemble $2,000-3,000, Multi-Custodian $2,800-4,000, OSJ adds $1,500-3,000/supervised advisor incremental, Wirehouse FA depends on home-office Year 3 strategy.

Marketing Rule and ADV implications of Year 3. The service-tier launch is the highest Marketing Rule exposure event in the three-year roadmap. Every claim — "AI-augmented planning," "agentic AI for operations," "AI-native service tier" — gets the substantiation file, the "clear and prominent" disclosure under Rule 206(4)-1, and the L4 Ch7 L1 audit. The ADV Part 2A is materially updated (L5 Ch7 off-cycle amendment). The engagement letter language is rewritten. The conflict-disclosure narrative around AI's effect on cost-to-serve (and the firm's margin) is addressed under the Care and Conflict Obligations of Reg BI §240.15l-1. M&A buyers value Year 3 maturity at the AI-maturity-premium top of the range — top-quartile RIAs at 8x-10x adjusted EBITDA per Mercer Capital / ECHELON Q3-Q4 2025, with documented agentic-AI maturity adding to the upper end and premium-top transactions reaching ~11.6x.

Three Failure Modes and the Defensible Roadmap

Three failure patterns kill most three-year roadmaps in advisor practices. Naming them up front is the practical inoculation.

Premature Year 3 ambition. The practice that buys agentic AI in Year 1 — usually because a vendor demo was compelling — without the workflow documentation, archive pipeline, vendor DD, WSP, or supervisory tooling Year 2 establishes. The agentic AI produces actions the practice cannot supervise, archive, or document. FINRA Rule 3110 reasonable-design becomes the examination finding. The remediation is to roll back the agentic deployment, complete the Year 1-2 fundamentals, and re-deploy on the proper foundation. Cost: 6-12 months of wasted work and a documented near-miss in the firm's incident history.

Year 1 tool sprawl. The solo or small ensemble that buys Jump, Zocks, FinMate, Sybill, Pulse360, Holistiplan, FP Alpha, Wealth.com, RightCapital, eMoney, MoneyGuidePro, Wealthbox, Redtail, Salesforce FSC, Practifi, Microsoft Copilot, OpenAI Enterprise, and Claude for Work — without picking one per category. Year 1 budget triples, advisor adoption fragments, no tool is mastered, and the L3 capstone workflows never operationalize. The remediation is ruthless consolidation per the L4 Ch1 L1 solo discipline (one per category).

Year 2 integration debt. The practice that buys Year 1 productivity tools but skips the integration year — never connects Jump to CRM, never builds RAG to firm vault, never operationalizes the L3 Ch10 archive pipeline. Year 1's promised time savings show up partially; advisors still manually re-key data; the archive misses 15-30% of artifacts; the supervisor cannot conduct Rule 2210 review at scale. Year 3 ambition becomes impossible because the foundation rotted in Year 2. Remediation: a year of "integration-only" investment with no new tools, which the CFO will resist and the advisors will perceive as standing still.

The defensible roadmap avoids all three by sequencing rigorously: Year 1 productivity with single-vendor-per-category discipline and L4 Ch2 vendor DD, Year 2 integration with the L3 capstone playbook as the deliverable and the L4 Ch3 WSPs and L4 Ch4 cyber program as the regulatory floor, Year 3 differentiation with proprietary workflows, agentic AI under Rule 3110, and AI-native service tiers — each gated by completion of the prior year's deliverables. The one-page roadmap document, signed by the managing partner and CCO, filed under the WSP, re-baselined annually after the readiness audit, becomes the practice's defensible answer to the regulator, the CFO, the buyer, and the next-generation advisor evaluating the firm.

Key Takeaways

• Three-year horizon, not five (too long), not one (too short). The AI, regulatory, vendor, and M&A landscapes all move at a three-year clock; the L4 chapters sequence to it.
• Year 1 — Productivity: meeting AI, planning AI, enterprise LLM, ADV Part 2A disclosure live, L4 Ch2 vendor DD per tool. Budget anchors $400-2,000/advisor seat/month all-in for the AI layer depending on archetype.
• Year 2 — Integration: CRM ingestion of structured AI output, archive coverage, RAG to firm vault, supervisory tooling, May 2024 Reg S-P IRP operationalized, L4 Ch3 WSPs live. Budget $700-2,800/advisor seat/month.
• Year 3 — Differentiation: proprietary workflows, agentic AI under Rule 3110 reasonable-design with the L4 Ch3 L3 WSP, AI-native service tiers driving ADV fee schedule and conflict-disclosure updates. Budget $1,000-4,000/advisor seat/month.
• Three failure modes to avoid: premature Year 3 ambition (agentic AI without supervisory foundation), Year 1 tool sprawl (multiple vendors per category), Year 2 integration debt (Year 1 tools never wired into the stack).
• Marketing Rule and ADV implications scale across years: Year 1 introduces AI tool disclosure; Year 2 deepens the RAG and supervisory tooling disclosure; Year 3 launches service tiers that materially update fee schedule, engagement letter, conflict-disclosure narrative, and substantiation file under Marketing Rule 206(4)-1 and the January 2026 staff FAQs.
• One-page roadmap, two signatures (managing partner + CCO), annual re-baseline after the L4 Ch1 L2 readiness audit, filed under WSP as compliance artifact under SEC Rule 204-2 and FINRA Rule 4511.
• M&A defensibility: documented three-year roadmap with measurable Year 1-2 deliverables and Year 3 ambition is part of the L4 Ch8 L2 diligence pack; supports the AI maturity premium attribute of +0.5-1.5x on the top-quartile 8x-10x adjusted EBITDA multiple per Mercer Capital / ECHELON Q3-Q4 2025, with premium-top reaching ~11.6x.