01
Execution becomes infrastructure.
Clouds are bundling runtime, identity, memory, tools, sandboxes, policy, observability and distribution.
Do not build where capex wins.
Actioneer product strategy / July 2026
The next layer is not agents.
It is the operating system that decides which work deserves authority, what it changed, and where the next dollar goes.
The wedge
Operated onboarding + cross-sell programs
Research finds the intervention. Voice executes it. Evidence decides what scales.
Thesis, not forecast. External facts and internal shipping evidence are separated throughout.STRATEGY
Start with forces, not features
Three things we cannot wish away.
02
Action follows data gravity.
Salesforce, ServiceNow, SAP and Oracle put agents beside the records, permissions and workflows they already own.
Integrate; do not recreate every system of record.
03
Accountability stays fragmented.
Each provider can meter its own activity. None is neutral about whether the customer should use less of it.
Own the cross-system decision record.
AWS AgentCore · Google Agent Platform · Salesforce Agentforce · Sierra Horizon, 16 Jul 2026FACT → INFERENCE
Force 01 / capital
The infrastructure contest has already been capitalised.
Alphabet capex 2025
$91.4B
2026 guidance
$175–185B
Company-reported. Large majority of 2026 capex directed to technical infrastructure.
Actioneer’s move
↑
Move up the stack.
Rent models, media, runtime and marketplaces. Own the program contract, authority, evidence and economic decision.
Alphabet 2025 Q4 earnings call, 4 Feb 2026COMPANY-REPORTED FACT
The stack
Every layer can be valuable. Only one is plausibly ours.
Program economicsobjective → authority → intervention → evidence → outcome → next allocationOWN
Research + executionActioneer Research · Voice · campaigns · operatorsBUILD SELECTIVELY
Systems of recordSalesforce · ServiceNow · PostHog · warehouses · CDPsADAPT
Agent platformsSierra · Lyzr · ElevenLabs · Ringg · NurixUSE / COMPETE AT JOB
Execution substrateAWS · Azure · GCP · models · speech · telephonyRENT
The choice is architectural and commercial: portability above providers, not parity beneath them.STRATEGIC CHOICE
Actor map
The ecosystem is crowded. The gap is still real.
Build + run agents
Sarvam · ElevenLabs · Ringg · Sierra · Nurix · Lyzr
Models, media, orchestration, builders, contact-centre agents, deployment.
See + understand
Hex · PostHog
Analytics, notebooks, product events, experiments, customer data.
Customer-aligned gap
Transform + operate
Accenture · EY · BCG X
Executive access, domain expertise, implementation capacity, managed services.
What none naturally owns
Cross-provider capital allocation
A durable, customer-owned record of why work was funded, what was authorised, what changed, and what scales next.
Game theory, not feature theory
Each incumbent is rational. That creates the opening.
Actor
Must maximise
Can credibly prove
Will not volunteer
Cloud / model
Consumption + ecosystem gravity
Runs, latency, cost, policy
“Use a rival—or use less.”
System of record
Actions inside its suite
Native activity + records
Cross-suite displacement.
SI / operator
Services revenue + client trust
Delivery + transformation
Commoditise its own domain IP.
Enterprise buyer
Incremental gross profit
Only with joined evidence
Needs an aligned allocator.
Inference from published business models and control points. “Neutral control plane” claims do not remove supplier economics.GAME-THEORETIC INFERENCE
Start from the actual buyer
The buyer is not buying “AI.” They own a number.
Primary buyer
GM · CRO / CGO · Head of Customer Operations
→
Their contract with us
Move onboarding, conversion, retention or cross-sell—within a committed budget and an auditable operating boundary.
- Fixed operating fee creates urgency
- Upside-linked fee aligns economics
- Evidence grade makes settlement defensible
Buyer and fee structure are Actioneer field learning from prior pilots; not an externally audited market statistic.FOUNDER / GTM EVIDENCE
Internal shipping evidence
The code is already converging on a program, not a chatbot.
22 Jun
Campaign success metrics + in-system attributionactioneer-voice · 3e4bbb7 / 1c425fb
22 Jun
Campaign guardrails + connectionsactioneer-voice · d97a731 / d7871eb
24–30 Jun
CDP activation workspace → mid-call WhatsApp → click attributionactioneer-voice · 9443ec3 / 330f89c
25 Jun–7 Jul
Segments, funnels, retention and cohort export shipped across product branchesresearch-analyst · branch history
9–10 Jul
Bancassurance cross-sell pool + Indic voice simulation and gradingbaby-sentinel · 09f0369b / 698016be
15–16 Jul
Shared Actioneer identity, shell and design language workresearch-analyst + baby-sentinel
Repository history proves product-shipping direction, not customer demand or commercial traction.VERIFIED LOCAL GIT HISTORY
The wedge
Sell one complete operating loop.
1. ObjectiveOnboarding / cross-sell KPI
→
2. ResearchWho, why, which intervention
→
3. AuthorityBudget, consent, permitted action
→
4. Voice + channelsProvider-backed execution
7. Next allocationScale, redirect, stop
←
6. ImpactCounterfactual + evidence grade
←
5. ReceiptsCost, action, human work, outcome
←
One managed programOne accountable owner
Thin vertical slice; broad destination. Voice is the first governed capability, not the category.PRODUCT CHOICE
Commercial grammar
A managed program is a three-part contract.
Operating agreement
What we run
Objective, scope, owner, SLA, fixed fee, approvals, operator responsibility.
+
Authority mandate
What may act
On-behalf-of, budget cap, data grants, channels, consent, expiry, revocation.
+
Impact evaluation
What earns upside
Baseline, counterfactual, outcome source, evidence threshold, caps, exclusions, dispute.
Separate operating work, causal evaluation and commercial settlement. Conflating them makes every ambiguity billable.DOMAIN MODEL
Architecture follows the business
One platform. Products that can still move independently.
SHARED ACTIONEER SHELL · identity · tenancy · connections · design language · observability
MANAGED PROGRAMS · contract · allocation · cases · settlement · portfolio
RESEARCH
VOICE
OPERATOR STUDIO
Evidence + receipts
Authority + approvals
DataGateway
Work + spend meter
Deep Agents / harnesses
Sarvam / ElevenLabs
Ringg / Nurix
CRM / CDP / PostHog
Clouds / sandboxes
SIs / domain operators
A monorepo does not require one deployment. Interfaces define release independence; folders do not.ARCHITECTURAL PRINCIPLE
What actually compounds
The moat is not the harness. It is the authorised history.
Context
→
Thesis
→
Authority
→
Spend
→
Action
Evidence
→
Outcome
→
Impact
→
Allocation
↻
Better canon
More runs
≠
More value
;
Better allocations = value
Cross-customer learning only with explicit rights and privacy-safe artifacts. Customer raw data need not become the moat.MOAT THESIS
Distribution + delivery
Scale programs through operators—then share the economics.
PROGRAM
CANON +
CONTROL PLANE
CANON +
CONTROL PLANE
Actioneer ships a proven ProgramKit
SI / domain expert operates it
Evidence improves the canon
More customers fund more learning
The franchise bargain
Partners keep vertical services revenue and upside share.
Actioneer provides the portable grammar: authority, evidence, evaluation, provider adapters, operator tooling and settlement.
$9.39B
Accenture Q3 managed-services revenue
$9.33B
consulting revenue
Not generic BPO
Accenture Q3 FY2026 · AWS partner-led FDE, 30 Jun 2026 (“delivery IP stays with the partner”)FILED FACT + OFFICIAL ANNOUNCEMENT
Explicit non-choices
The choices make the strategy.
×
No foundation-model or voice-media stack.Use Sarvam, ElevenLabs and others behind a stable capability contract.
×
No horizontal CX Agent OS.Sierra has chosen that game, with capital, field teams and persistent conversation data.
×
No generic agent control tower.Clouds, Microsoft and ServiceNow will bundle governance into their estates.
×
No warehouse, notebook or product-analytics rebuild.Integrate Hex, PostHog and customer systems as evidence/action providers.
×
No partner-specific forks.Partners extend ProgramKits and adapters; the platform remains one product.
×
No pure outcome pricing.A fixed operating fee funds committed execution; upside pays for auditable impact.
“Voice AI”, “AI analytics”, “agent platform” and “consulting” are all adjacent capabilities—not our chosen category.STRATEGIC BOUNDARY
Sequence without descoping
Two months proves the loop. Five years changes who allocates software labour.
0–2 months
One deployable program
- Research foundation
- Onboarding / cross-sell
- Provider-backed Voice
- Receipts + human gate
3–12 months
Repeatable operation
- Fixed + upside settlement
- Operator surface
- ProgramKit canon
- Second operator deploys without core changes
1–5 years
Authority, evidence + capital plane
- Cross-provider portfolios
- Customer-boundary deployment
- Partner-operated verticals
- Marginal allocation by evidence grade
- Software labor treated as an investable operating resource
The end-state is conditional. Product milestones should earn—not assume—the right to broaden scope.SEQUENCING
The thesis is asymmetric, not unquestionable
Three truths we must earn quickly.
?
Will a buyer grant durable access to the evidence needed to evaluate a program?
Test: one lighthouse customer supplies baseline, actions, outcome source and correction path.
?
Does the second deployment get materially faster and more profitable?
Test: a non-core operator launches a ProgramKit without changing platform code.
?
Will evidence change the next budget decision—not just decorate a report?
Test: capital is scaled, redirected or stopped because of an explicit evidence grade.
If attribution remains weak, the fallback is still valuable: an evidence ledger and better human allocation—not fabricated autonomy.DISCONFIRMATION TESTS
The five-year destination
The accountable growth machine.
Actioneer operates AI-native revenue programs—and proves which interventions deserve the next dollar.
Today: Research + Voice.
Then: Programs + operators.
Ultimately: the customer-aligned authority, evidence and capital layer for software labor.
Then: Programs + operators.
Ultimately: the customer-aligned authority, evidence and capital layer for software labor.
Sell the operation. Productise the operating grammar. Let every program make the next one easier to trust, deploy and fund.ACTIONEER