Skip to content
02 · Insights— The analysis

Insights

What 3,891 deployments add up to. Six questions, each answered from the full record and recomputed as new deployments arrive. No survey panels, no opinions.

Data updated

  1. Q1Where is AI actually deployed?Healthcare532 deployments, the most of any industry
  2. Q2Which business functions does it serve?Operationstouched by 86% of deployments
  3. Q3Which patterns keep recurring?230 patternsled by customer service automation
  4. Q4What results do deployments report?−50%median reported time reduction
  5. Q5Is it built, bought or low-code?72%custom-built, where the approach is known
  6. Q6How agentic is enterprise AI?28.9%of deployments use AI agents
§ 1

Where is AI actually deployed?

Healthcare leads with 532 deployments (13.7%). The top three industries hold 35% of the record.

Each deployment belongs to exactly one industry, so these shares add up to 100%.

  1. Healthcare13.7% of the record
    532
  2. Finance10.9% of the record
    425
  3. Tech & Comms10.3% of the record
    402
  4. Manufacturing10% of the record
    391
  5. Professional Services6.3% of the record
    247
  6. Insurance6.3% of the record
    244
  7. Retail6.1% of the record
    237
  8. Public Sector5.3% of the record
    207
8 of 18 industries
§ 2

Which business functions does it serve?

Operations is involved in 86% of deployments; Sales in 12%.

One deployment can serve several functions, so shares do not add up to 100%.

  1. Operations3,347 deployments
    86%
  2. Legal & Compliance1,663 deployments
    43%
  3. Human Resources1,641 deployments
    42%
  4. Research & Development1,463 deployments
    38%
  5. Supply Chain1,213 deployments
    31%
  6. Customer Service1,034 deployments
    27%
  7. Finance915 deployments
    24%
  8. Marketing870 deployments
    22%
  9. IT & Security505 deployments
    13%
  10. Sales449 deployments
    12%
§ 3

Which patterns keep recurring?

Customer service automation is the most documented pattern: 212 deployments, 52 of them published in the last 12 months.

Browse every pattern
  1. Customer service automationMost reported: Customer experience & trust
    212
  2. AI platformMost reported: New product / capability
    133
  3. Workflow automationMost reported: Speed & agility
    131
  4. Predictive maintenanceMost reported: New product / capability
    126
  5. Claims automationMost reported: Speed & agility
    105
  6. Agriculture optimizationMost reported: New product / capability
    87
  7. Patient engagementMost reported: Customer experience & trust
    81
  8. Clinical documentationMost reported: New product / capability
    80
  9. Risk assessmentMost reported: Risk & compliance
    78
  10. Compliance automationMost reported: Risk & compliance
    76
10 of 230 patternsAll 230 patterns
§ 4

What results do deployments report?

1,428 deployments (36.7%) publish at least one quantified result. Reported time reductions have a median of 50%.

Medians describe what sources claim. Many sources are customer stories published by vendors.

  1. Time & speed

    1,179 metrics · 814 deployments

    Reduction · median and middle half
    −50%Median reduction −50%, middle half −30% to −76%, from 283 metrics
    Increase · median+59%128 metrics

  2. Cost savings

    278 metrics · 241 deployments

    Reduction · median and middle half
    −40%Median reduction −40%, middle half −25% to −70%, from 168 metrics
    Increase · median+47%11 metrics

  3. Productivity & throughput

    145 metrics · 130 deployments

    Reduction · median and middle half
    −50%Median reduction −50%, middle half −46% to −51%, from 9 metrics
    Increase · median+40%91 metrics

  4. Quality & accuracy

    101 metrics · 91 deployments

    Reduction · median and middle half
    −30%Median reduction −30%, middle half −20% to −49%, from 21 metrics
    Increase · median+40%50 metrics

  5. Revenue & growth

    109 metrics · 87 deployments

    Reduction · median and middle half
    −31%fewMedian reduction −31%, middle half −29% to −33%, from 2 metrics
    Increase · median+40%67 metrics

  6. Automation & deflection

    84 metrics · 75 deployments

    Reduction · median and middle half
    −50%Median reduction −50%, middle half −40% to −73%, from 23 metrics
    Increase · median+49%19 metrics

  7. Adoption & scale

    94 metrics · 73 deployments

    Reduction · median and middle half
    −56%fewMedian reduction −56%, middle half −39% to −73%, from 2 metrics
    Increase · median+250%8 metrics

  8. Customer experience

    65 metrics · 55 deployments

    Reduction · median and middle half
    −30%Median reduction −30%, middle half −28% to −58%, from 8 metrics
    Increase · median+28%50 metrics

  9. Risk, reliability & safety

    42 metrics · 38 deployments

    Reduction · median and middle half
    −70%Median reduction −70%, middle half −60% to −90%, from 13 metrics
    Increase · median+70%13 metrics

  10. Sustainability & resources

    27 metrics · 24 deployments

    Reduction · median and middle half
    −25%Median reduction −25%, middle half −20% to −30%, from 20 metrics
    Increase · median+30%1 metric · few

  11. Other quantified impact

    628 metrics · 442 deployments

    Reduction · median and middle half
    −40%Median reduction −40%, middle half −25% to −62%, from 102 metrics
    Increase · median+50%161 metrics

Grey figures rest on fewer than 5 published metrics. Read them as examples, not as a typical result.

How these figures are measured
Every quantified claim in a source is extracted, assigned to one outcome category and, where it can be expressed as a percent change, normalized to a reduction or an increase. The bar shows the median and the middle half (25th to 75th percentile) of reductions. Increases are shown as a median only, because they are unbounded. Deployments without a published number are excluded, not counted as zero.
§ 5

Is it built, bought or low-code?

Custom code dominates: 72% of deployments with a known approach were built rather than bought.

The build approach is known for 2,800 of 3,816 deployments (73%). The rest are left out.

  • Custom-built72% · 2,026
  • Low-code10% · 291
  • Bought9% · 247
  • Mixed8% · 236
§ 6

How agentic is enterprise AI?

28.9% of deployments use AI agents, and 37.5% of those coordinate several agents. In the last 12 months the share is 33.4%, against 32.8% before.

Each function compares its agent share in the last 12 months with deployments published from 2023 up to a year ago. ≈ marks a change a 95% test cannot tell from chance.

See how the share is changing
1,075
agentic deployments
of 3,726 classified
403
multi-agent systems
several agents working together
46%
agentic in IT & Security
highest share of any function, last 12 months

Agent share by business function, last 12 months

  1. IT & Security97 of 210 · ≈ +6.8 pts vs. before
    46%
  2. Sales50 of 119 · ≈ −2.2 pts vs. before
    42%
  3. Finance95 of 229 · +14 pts vs. before
    42%
  4. Legal & Compliance183 of 467 · ≈ +3.8 pts vs. before
    39%
  5. Research & Development151 of 407 · ≈ +5.3 pts vs. before
    37%
  6. Operations353 of 974 · ≈ +1.8 pts vs. before
    36%
  7. Supply Chain110 of 316 · ≈ +1.2 pts vs. before
    35%
  8. Customer Service104 of 299 · −8.1 pts vs. before
    35%
  9. Human Resources125 of 410 · ≈ +0.1 pts vs. before
    31%
  10. Marketing95 of 313 · ≈ −6.8 pts vs. before
    30%

Every number on this page is computed from individual records. Open any bar to reach the dossier, and from there the deployments behind it.

Browse the evidenceMethodology