The Three-Horizon AI Portfolio:
60/30/10 Quick Wins,
Advanced, and Experimental
Single-bet AI strategies fail. Portfolio thinking wins. The three-horizon framework from Chapter 13 of The AI Strategy Blueprint tells CIOs and Chief AI Officers exactly how to allocate AI investment across proven productivity wins, advanced applications, and experimental bets — and how to rebalance as the technology evolves.
- Single-project AI thinking creates the pilot purgatory trap. Portfolio thinking — managing AI investments across three time horizons simultaneously — is the antidote.
- Horizon 1 (60–70%): Proven technologies and immediate quick wins. Chat assistants, document summarization, meeting intelligence, routine content generation. Deploy these first; they build literacy and fund the rest.
- Horizon 2 (20–30%): Established AI with customization. RAG systems, workflow automation, domain-specific applications, bounded agentic AI. Funded by demonstrated Horizon 1 ROI.
- Horizon 3 (10–20%): Experimental bets on emerging technology. Full agentic workflows, multimodal applications, industry-specific advanced models. Small allocation, high optionality value.
- Rebalance quarterly — not weekly. As technologies mature, items migrate from H3 → H2 → H1. Gartner's 33% agentic AI projection for 2028 is the proof that today's H3 bets become tomorrow's H1 requirements.
- Why Portfolio Thinking Beats Single-Bet AI
- Horizon 1: Quick Wins (60–70%)
- Horizon 2: Advanced (20–30%)
- Horizon 3: Experimental (10–20%)
- The 60/30/10 Allocation Logic
- The Gartner 33% Agentic by 2028 Projection
- Build vs Buy vs Partner Matrix
- Rebalancing the Portfolio Quarterly
- Related Case Studies
- Frequently Asked Questions
Why Portfolio Thinking Beats Single-Bet AI
"Organizations that overweight Horizon 3 while neglecting Horizon 1 fall into the perpetual pilot trap." This observation from Chapter 13 of The AI Strategy Blueprint identifies one of the two most common AI strategy failure modes. The other is the mirror image: organizations that deliver Horizon 1 quick wins without any Horizon 3 experimentation find themselves perpetually behind as the technology landscape shifts.
The portfolio mental model solves both failure modes simultaneously. By design, it requires that 60–70% of AI investment be in proven technologies delivering measurable returns — ensuring that AI programs fund themselves through early ROI. And by design, it requires that 10–20% of investment go to experimental bets — ensuring that organizations are building competency in technologies that will become mainstream within 24–36 months.
"AI is not a single technology but a taxonomy of distinct capabilities, each suited to different problem types. Successful organizations avoid treating every challenge as an LLM problem, match technology to the problem, and balance their portfolio across three horizons."
— John Byron Hanby IV, The AI Strategy Blueprint, Chapter 13
The financial logic reinforces the strategic logic. Horizon 1 investments in chat assistants, document summarization, and meeting intelligence typically deliver productivity gains of 10–30% for knowledge workers within 30–60 days of deployment. Those gains create organizational credibility for AI investment — and generate the internal political capital to fund Horizon 2 and Horizon 3 bets.
The pilot purgatory article addresses why single-project AI thinking creates the specific failure mode where organizations run multiple pilots indefinitely without graduating to production. The portfolio framework is the antidote at the program level: instead of evaluating each AI initiative in isolation, evaluate the full portfolio balance and fund accordingly.
Horizon 1: Quick Wins (60–70%)
The majority of your AI investment — 60 to 70 percent — belongs in Horizon 1: proven technologies that deliver immediate productivity gains. This is not a concession to conservatism. It is a recognition that quick wins fund the rest of the portfolio, build the organizational AI literacy that determines long-term success, and create the credibility required for Horizon 2 and 3 investment.
AI Chat Assistants for Every Knowledge Worker
The single highest-ROI first deployment for most organizations. Local AI chat assistants — AirgapAI is the reference implementation — enable any employee to draft emails, summarize documents, generate content, and answer questions without technical expertise. Deployed to 100% of the workforce via one-click installer. No integration required. Value delivered in hours.
Document Summarization
Long reports, contracts, research papers, policy documents, and technical manuals — all summarizable in seconds. A 16-page contract that requires 30 minutes of careful reading can be analyzed in seconds. For organizations managing high document volumes, this single use case frequently justifies the entire Horizon 1 deployment cost.
Meeting Intelligence
Automated transcription, action item extraction, and meeting summaries distributed to attendees within minutes of call completion. For organizations with high meeting cultures — professional services, enterprise sales, program management — this use case alone recovers 3–5 hours per week per employee in administrative overhead.
Routine Content Generation
Status updates, internal communications, first-draft proposals, social media posts, and email sequences — generated at 10x the speed of manual drafting. The employee retains editorial control; the AI handles the blank-page problem and structural scaffolding.
Chapter 13 is explicit about the literacy dimension of Horizon 1: "These applications use established technology, require minimal customization, and deliver immediate productivity gains." The implicit benefit is equally important — every Horizon 1 deployment builds the AI literacy that the 10-20-70 rule identifies as the primary success factor for AI transformation.
Horizon 2: Advanced (20–30%)
20 to 30 percent of AI investment belongs in Horizon 2 — established AI technologies with customization that deliver organizational-level impact rather than individual-level productivity. Horizon 2 applications require more investment than Horizon 1 but deliver proportionally greater business outcomes when use cases are properly validated.
RAG Systems for Organizational Knowledge Bases
Retrieval-Augmented Generation (RAG) grounds AI responses in your organization's actual documents — eliminating the hallucination risk that blocks production deployment for high-stakes use cases. Policy Q&A, technical documentation, contract analysis, and call center knowledge bases all belong here.
The quality of RAG outputs depends critically on data preparation. The RAG vs fine-tuning analysis explains why intelligent data ingestion — not model size — determines accuracy. The 78x accuracy improvement that Blockify delivers over naive chunking represents the difference between production-ready RAG and a system that hallucinates on 20% of queries.
Workflow Automation for High-Volume Processes
Platforms like n8n connect AI with 500+ business applications to create end-to-end automated processes. A marketing team configuring a workflow that monitors a content calendar, triggers AI generation, saves output to Google Drive, and updates the originating spreadsheet — without coding. One technology company used this approach to generate 100+ SEO-optimized articles in a single weekend.
Workflow automation requires specialized expertise to configure reliably. Organizations should designate skilled practitioners who translate business requirements into production workflows — not every employee will build their own automations.
Agentic AI for Bounded Multi-Step Workflows
Agentic AI — systems that pursue goals with environmental awareness, planning capability, and tool use — belongs in Horizon 2 when deployed in bounded, well-defined contexts with human oversight. Sales qualification pipelines, research synthesis workflows, and data extraction pipelines are representative Horizon 2 agentic deployments.
Full autonomous agentic AI for complex, unbounded tasks belongs in Horizon 3 until the technology matures further. The distinction: Horizon 2 agentic AI operates within defined boundaries with human review at critical steps; Horizon 3 agentic AI experiments with greater autonomy.
Horizon 3: Experimental (10–20%)
The 10–20% allocation to Horizon 3 is not waste — it is optionality. Organizations that have no Horizon 3 investment find themselves perpetually in reaction mode as AI capabilities evolve. By the time a technology matures to enterprise readiness, they are starting from zero; competitors who invested in Horizon 3 are moving to Horizon 1.
"Organizations that defer AI adoption waiting for 'better' models will find themselves perpetually waiting while competitors capture value with current technology."
— The AI Strategy Blueprint, Chapter 13
Current Horizon 3 investment areas (as of 2026):
- Fully autonomous agentic workflows — complex multi-step tasks that currently require significant human oversight. Autonomous contract negotiation support, cross-system data reconciliation, autonomous research tasks with complex decision branching. The Gartner 33% projection for 2028 signals that today's H3 becomes tomorrow's H2 rapidly.
- Multimodal AI applications — combining text, image, and video for quality inspection, document processing with visual elements, and video-based training and simulation. The technology is maturing rapidly but enterprise-grade tooling is still emerging.
- Industry-specific advanced model applications — drug discovery in pharma, satellite imagery analysis in defense, protein folding in life sciences. High potential, high uncertainty, small allocation.
- AI-to-AI orchestration — systems where AI models orchestrate other AI models for complex compound tasks. Early-stage enterprise applicability but significant strategic value for organizations that learn the patterns now.
The governance principle for Horizon 3 is different from Horizon 1 and 2: learning is the deliverable. Success is not production deployment — it is organizational knowledge about what works, what does not, and what the path to production requires. Set that expectation in budget conversations before funding Horizon 3 work.
The 60/30/10 Allocation Logic
The allocation percentages are not arbitrary. Chapter 13 of The AI Strategy Blueprint grounds the 60/30/10 ratios in the relationship between organizational AI literacy, technology maturity, and risk tolerance at each stage of an AI program.
The 60–70% Horizon 1 floor ensures that AI programs are self-funding: enough value is delivered quickly enough to justify continued investment. It also ensures that AI literacy — the skill distribution across the workforce that determines long-term AI program success — builds through actual usage rather than training programs alone.
The 20–30% Horizon 2 allocation creates the pathway from individual productivity to organizational impact. Horizon 2 applications — RAG systems, workflow automation, bounded agentic AI — require the Horizon 1 literacy base to be effective. Employees who have been using AI assistants for six months understand how to query AI systems, set appropriate expectations, and identify the failure modes that require human oversight.
The 10–20% Horizon 3 ceiling is as important as the floor. Organizations that over-allocate to experimental AI — a common failure mode in technology-forward cultures — create two problems: they starve the Horizon 1 deployments that fund the program, and they build organizational exposure to high-risk bets without the risk diversification that portfolio construction provides.
For organizations at the beginning of their AI journey — fewer than 12 months in production — the allocation should weight toward 70% Horizon 1, 25% Horizon 2, and 5% Horizon 3. As organizational AI literacy grows and Horizon 1 ROI is established, the allocation can shift toward the 60/30/10 steady-state target. The AI change management framework provides the people-side structure that supports this progression.
The AI Strategy Blueprint
Chapter 13 of The AI Strategy Blueprint contains the complete AI taxonomy (traditional ML, generative AI, agentic AI), the parameter-to-capability table, the RAG vs fine-tuning decision framework, the build/buy/partner matrix, and the three-horizon portfolio strategy with allocation guidance. Available on Amazon.
The Gartner 33% Agentic by 2028 Projection
Gartner projects that by 2028, 33% of enterprise software will include agentic AI — up from less than 1% today. This single data point is the most important input to Horizon 3 portfolio allocation decisions.
The implications cascade through the entire portfolio framework:
- Today's Horizon 3 (experimental agentic AI) will migrate to Horizon 2 (established with customization) within 18–24 months. Organizations without current Horizon 3 agentic investment will be starting Horizon 2 agentic deployment from zero when their competitors are scaling it.
- The 33% penetration rate means that by 2028, most enterprise software vendors — ERP, CRM, HCM, supply chain — will embed agentic capabilities into their products. Organizations building internal AI competency now will be positioned to leverage those vendor capabilities rather than learn from scratch.
- The transition from 1% to 33% in four years is historically fast for enterprise technology adoption. The only comparable speed was the adoption of cloud SaaS from 2010–2015. Organizations that missed the SaaS transition spent years playing catch-up. The agentic AI transition will be steeper.
"Gartner predicts that by 2028, 33% of enterprise software will include agentic AI, a dramatic increase from less than 1% today."
— The AI Strategy Blueprint, Chapter 13, citing Gartner
The practical implication for portfolio allocation: Horizon 3 agentic investment is not optional for organizations that want to lead. The 10–20% allocation is calibrated for risk management — small enough that failures are affordable, large enough that the learning is meaningful. Start now. The crawl-walk-run framework applies at the individual use case level; the three-horizon portfolio applies at the program level.
Build vs Buy vs Partner Matrix
Each horizon of the AI portfolio has a natural fit with one of three development approaches — build, buy, or partner. Chapter 13 maps these relationships explicitly and provides the decision criteria that determine when each approach is appropriate.
Buy
Advantages: Fastest deployment, vendor expertise, proven quality, reduced maintenance burden. Off-the-shelf chat assistants, meeting intelligence tools, and document summarization products deliver Horizon 1 value without engineering investment.
Best for: Common use cases where speed matters, organizations with limited AI engineering capacity, and any deployment where a vendor's existing product already solves the problem adequately.
Horizon 1 examples: AirgapAI for chat assistant deployment, Otter.ai for meeting intelligence, existing document management platform AI features.
Partner
Advantages: Access to specialized expertise, shared investment in solution development, capability transfer over time, risk sharing on complex initiatives.
Best for: RAG system implementation, workflow automation that requires deep integration with enterprise systems, and any Horizon 2 application where internal teams lack the specific AI engineering competency.
Horizon 2 examples: Consulting engagements for RAG platform design, SI partnerships for workflow automation, the Iternal AI Strategy Sprint for full architecture and deployment.
Build
Advantages: Maximum customization, genuine competitive differentiation, full control over architecture and roadmap.
Best for: Capabilities that represent core competitive differentiation — use cases where the AI application itself is the product, where vendor solutions do not exist, or where proprietary data creates an advantage that only a custom build can capture.
Horizon 3 examples: Custom agentic AI orchestration for proprietary processes, industry-specific models trained on proprietary operational data, AI-native products where the AI is the differentiator.
Most organizations deploy all three approaches across their AI portfolio. The decision for each initiative depends on strategic importance, internal capability, time constraints, and available vendor solutions. The framework in Chapter 13 avoids the common failure of applying a single approach to all AI initiatives — treating everything as "buy" creates dependency and misses differentiation opportunities; treating everything as "build" creates engineering bottlenecks and misses time-to-value.
Rebalancing the Portfolio Quarterly
"Plan for quarterly model evaluations, not weekly updates." This guidance from Chapter 13 applies equally to portfolio rebalancing: the cadence should match the pace at which meaningful changes occur in the AI technology landscape and in organizational AI maturity.
Quarterly rebalancing has three components:
Evaluate ROI Across Active Deployments
Review measured outcomes for all Horizon 1 and Horizon 2 deployments. Applications that have delivered validated ROI should be maintained and potentially scaled. Applications that have not shown measurable returns after 60–90 days should be evaluated for Scale / Iterate / Pivot / Stop — the four outcomes from the pilot evaluation framework.
Migrate Technologies Across Horizons
As AI technologies mature, they migrate: Horizon 3 experiments that prove viable move to Horizon 2 planning; Horizon 2 applications that achieve production readiness move to Horizon 1 scaled deployment. Track which technologies crossed the enterprise-readiness threshold during the quarter — the Gartner agentic AI trajectory is the benchmark signal for this migration rate.
Adjust New Project Funding to Rebalance
Do not disrupt live deployments to rebalance. Instead, use new project funding allocation to correct drift. If the portfolio has drifted to 50% Horizon 1, 40% Horizon 2, and 10% Horizon 3 — allocate more new project funding to Horizon 1 and less to Horizon 2 until the ratio returns to 60/30/10 at steady state. This approach preserves continuity in live applications while correcting strategic drift in the investment pipeline.
The model evolution dimension of rebalancing deserves specific attention. Chapter 13 notes that "LLaMA 3's 1-billion parameter model achieves equivalent benchmark performance to LLaMA 2's 13-billion parameter model, representing a 13x reduction in size while maintaining quality." Applications designed around specific model sizes and hardware requirements should be reviewed quarterly to determine whether efficiency improvements have made previously infeasible use cases viable. This regularly pulls Horizon 3 bets into Horizon 2 territory.
For organizations building out their AI governance infrastructure alongside their portfolio, the AI governance framework provides the oversight structure that applies to portfolio-level decisions, not just individual use cases. See also the AI transformation roadmap for the full 12-month planning horizon that the three-horizon framework fits within.
Three-Horizon Portfolio in Practice: Case Studies
Real deployments from the book — quantified outcomes from Iternal customers across regulated, mission-critical industries.
Enterprise Agility
An enterprise agility consulting firm deployed the three-horizon framework across its practice: Horizon 1 for client deliverable drafting, Horizon 2 for automated research synthesis, Horizon 3 for agentic client analysis.
- 14 distinct AI use cases identified and prioritized across all three horizons
- Horizon 1 delivered measurable ROI within 30 days of deployment
- Horizon 3 experiments running in parallel without disrupting core delivery
Global IT Services — Sales Enablement
A global IT services firm used the three-horizon portfolio to sequence AI deployment across its 10,000+ seller organization — quick wins for individual sellers, advanced RAG for deal support, experimental agentic for pipeline management.
- Horizon 1 chat assistants deployed to full seller organization in 6 weeks
- Horizon 2 deal-support RAG system reduced proposal creation time by 65%
- Portfolio allocation model adopted as company-wide AI governance standard
Top 3 Pharmaceutical
A top-three pharmaceutical company structured its AI program across all three horizons — from immediate regulatory document summarization (H1) to clinical trial data analysis RAG (H2) to experimental AI-assisted drug discovery (H3).
- Regulatory document review time reduced by 70% via Horizon 1 deployment
- Horizon 2 RAG system for trial data deployed with full audit trail compliance
- Horizon 3 skunkworks delivering competitive intelligence ahead of adoption curve
Build Your Team's AI Portfolio Management Skills
The Iternal AI Academy includes dedicated modules on AI portfolio strategy, the three-horizon framework, build/buy/partner decision-making, and the change management required to execute portfolio transitions. Start for $7/week.
- 500+ courses across beginner, intermediate, advanced
- Role-based curricula: Marketing, Sales, Finance, HR, Legal, Operations
- Certification programs aligned with EU AI Act Article 4 literacy mandate
- $7/week trial — start learning in minutes
AI Portfolio Strategy Consulting
Map your organization's current AI initiatives across the three horizons, identify gaps and over-allocations, and build the quarterly rebalancing governance that keeps your portfolio on track — with expert guidance from the team that wrote the book.
Frequently Asked Questions
The three-horizon AI portfolio framework, from Chapter 13 of The AI Strategy Blueprint, structures an organization's AI investments across three time horizons: Horizon 1 (60–70% of portfolio) for proven technologies and immediate productivity wins; Horizon 2 (20–30%) for established AI with customization that delivers greater impact; and Horizon 3 (10–20%) for emerging and experimental technologies that position the organization for future capability. The allocation percentages are deliberate — they prevent the two most common failure modes: over-indexing on experiments while neglecting immediate value, and over-indexing on quick wins while missing transformational capability.
Horizon 1 (60–70% of portfolio) focuses on proven technologies with immediate productivity returns. From Chapter 13 of The AI Strategy Blueprint: AI chat assistants deployed to all employees (the single highest-ROI first deployment for most organizations); document summarization for long reports, contracts, and research materials; meeting intelligence for automated notes, action items, and summaries; and routine content generation for emails, status updates, and communications. These applications use established technology, require minimal customization, deliver measurable productivity gains within weeks of deployment, and build the organizational AI literacy that the 10-20-70 rule identifies as the primary determinant of AI success.
Horizon 2 (20–30% of portfolio) advances to established AI with customization for greater organizational impact. From Chapter 13: RAG (Retrieval-Augmented Generation) systems for organizational knowledge bases — policy Q&A, technical documentation, contract analysis; workflow automation using platforms like n8n for end-to-end AI-powered processes; domain-specific AI applications built on validated patterns for sales enablement, HR, finance, and legal; and agentic AI for bounded multi-step workflows with human oversight. Horizon 2 requires more investment than Horizon 1 but delivers proportionally greater impact when use cases are properly validated through Horizon 1 deployment.
Horizon 3 (10–20% of portfolio) invests in emerging technologies that position the organization for future competitive capability. From Chapter 13: fully autonomous agentic AI for complex multi-step workflows that currently require significant human oversight; multimodal applications combining text, image, and video as the technology matures for enterprise use; industry-specific AI applications built on advanced models; and next-generation RAG architectures that go beyond current chunking limitations. Horizon 3 investments carry higher failure rates — that is expected and acceptable within the 10–20% allocation. The goal is organizational capability-building and optionality, not near-term ROI.
Gartner projects that by 2028, 33% of enterprise software will include agentic AI — a dramatic increase from less than 1% today. This projection, cited in Chapter 13 of The AI Strategy Blueprint, has direct implications for the three-horizon portfolio: the category that currently belongs in Horizon 3 (experimental agentic AI) will migrate to Horizon 2 (established with customization) over the next two to three years as the technology matures and enterprise tooling improves. Organizations that begin building agentic AI competency today through Horizon 3 experimentation will be positioned to accelerate adoption as the technology crosses the enterprise-readiness threshold — while organizations waiting for maturity will be starting from zero.
Chapter 13 of The AI Strategy Blueprint recommends quarterly portfolio evaluation — not weekly model tracking, but not annual budget reviews either. Quarterly cadence aligns with the pace at which: new AI capabilities cross the enterprise-readiness threshold (moving items from Horizon 3 to Horizon 2 or 2 to 1); use cases prove or fail to prove ROI (determining whether Horizon 1 deployments should scale or be discontinued); and organizational AI literacy evolves (changing the risk tolerance for Horizon 2 and 3 investments). At each quarterly review, evaluate allocation drift against the 60/30/10 target and adjust new project funding rather than disrupting live deployments.