UNCLASSIFIED // OSINT // ANALYST EYES ONLY
08FORECASTS
NO FILTER
UTC 2026-05-12 03:20:15Z
SYNC OK
HORIZON2026
HORIZON SYNC // 2026Active horizon auto-selected from the timeline scrubber. Click a tab to override.

HORIZON SCRUBBER

Agent maturation, real but uneven

CONF.
70
▲ EVIDENCE

Gartner: 40% of enterprise apps embed task-agents by EOY '26

▼ COUNTER-EVIDENCE

S&P: 31% in prod, 88% pilots fail; Gartner: >40% projects cancelled by '27

On-device AI inflection

CONF.
78
▲ EVIDENCE

iOS 27 'Extensions' lets users pick model; ~60% new PCs NPU-equipped

▼ COUNTER-EVIDENCE

Privacy & UX still inconsistent across vendors

Reasoning models do NOT commoditise on price

CONF.
65
▲ EVIDENCE

GPT-5.5 doubled to $5/$30; Opus 4.7 holds. 'Structured financial product.'

▼ COUNTER-EVIDENCE

DeepSeek V4 matches at 10–13× lower

Voice becomes primary interface

CONF.
80
▲ EVIDENCE

55% consumers; voice AI $22.5B '26; ~25% of YC W26 voice-first

▼ COUNTER-EVIDENCE

Only 29% of cos. deployed voice — deployment gap

Memory & personalisation as moat

CONF.
60
▲ EVIDENCE

Mem0 canonical 'memory as product'; benchmark suite emerging

▼ COUNTER-EVIDENCE

Privacy & portability concerns

SOURCES: GARTNER · IDC · MENLO · MCKINSEY · BCG · BAIN · SACRA · STANFORD HAI · IEA · OECD
DOC ID GAI-26-05 · COPY 001 OF 001
UNCLASSIFIED // OSINT // ANALYST EYES ONLY