Document-driven
The deliverable is a document: a contract, a filing, a due diligence report.
ABL 51st Conference·Mexico City·October 2026
Leading Through
Disruption
How technology is changing business, leadership and decision-making.
Adolfo Rangel Prado · Independent Consultant & Associate Professor, Universidad Panamericana
Today · 35 minutes
01Intelligence is becoming abundant
02It’s worth what you can verify
03What becomes possible
04Why faster lawyers don’t make better firms
05Delegate the work, keep the authority
06The firm that learns
07Leading through it
01Section
01Intelligence is becoming abundant
02It’s worth what you can verify
03What becomes possible
04Why faster lawyers don’t make better firms
05Delegate the work, keep the authority
06The firm that learns
01Intelligence is becoming abundant
What “abundant” looks like
The Navier–Stokes equations describe how water, air and blood move. Mathematicians have worked on them for ninety years.
≈10,000AI agents working in parallel
88 hof continuous work
Not one brilliant assistant.
Ten thousand of them, at once.
One expert10,000 agents
01Intelligence is becoming abundant
When intelligence is available to everyone, what stops being worth paying for?
A first-year associate and a senior partner can use the same model, from the same phone.
“If you value intelligence above all other human qualities, you’re gonna have a bad time.”
Ilya Sutskever co-founder of OpenAI · 2023
Judgment doesn’t win by default. It now has to prove what it adds.
02Section
01Intelligence is becoming abundant
02It’s worth what you can verify
03What becomes possible
04Why faster lawyers don’t make better firms
05Delegate the work, keep the authority
06The firm that learns
02It’s worth what you can verify
Anthropic research · 20 Oct 2026
An AI designed proteins against scorpion toxins. Then a lab checked.
- 318,484designs generated by the model
- 211tested in a physical lab, by two outside companies
- 17bound the toxin they were designed against
Binding, not neutralization. No antivenom is claimed.
Leiurus quinquestriatus under UV light · photo: Iskra.piotr, CC BY-SA 4.0
02It’s worth what you can verify
Intelligence is worth what you can verify
Where is the test?
MathematicsA proof that others can check
BiologyA wet-lab assay: 211 tested, 17 worked
LawSources, facts, review, and a decision you can defend
So: what is your firm’s lab?
02It’s worth what you can verify
Why law is first in line
Rules-bound
The rules are written down: statutes, regulation, the firm’s own playbook.
Verifiable
The output can be checked against a source. That is what makes errors catchable at scale.
The fields where AI moves fastest are the ones where work can be read, ruled and checked.
Framework: Mark Ajzenstadt, “The AI-native services firm”, Aug 2026. He sells AI services: read it as a lens, not a verdict.
02It’s worth what you can verify
What has already happened in legal
OpenAI · Sep 2026
Astra for Law
54% fully correct on hard legal research questions, vs 38.7% for the general model.
OpenAI’s own 200-question test
Harvey · Aug 2026
Tenet
A legal model built on an open base model. Top score on contract tasks at under ¼ of the cost.
Reported by Harvey on the LAB benchmark
ABC Legal · Aug 2026
50+ agents in production
About 50% lower cost on certain legal tasks. Fifteen non-engineers had agents running in a week.
Published by Anthropic about its customer
Read vendor numbers the way you read the other side’s exhibits: useful, and never neutral.
03Section
01Intelligence is becoming abundant
02It’s worth what you can verify
03What becomes possible
04Why faster lawyers don’t make better firms
05Delegate the work, keep the authority
06The firm that learns
03What becomes possible
From a new capability to a new service
01A new capabilityRead and compare every document, not a sample
02Work that was too expensiveThe whole data room. Every contract, not only the big ones
03A different serviceContinuous follow-up instead of a point-in-time opinion
04A firm built to deliver itWho checks, who decides, who answers to the client
A benchmark only proves the first link. Demand, quality and cost decide the other three.
03The question for this room
What service could your firm offer if certain intellectual capabilities were available continuously, and what would have to change to deliver it well?
03What becomes possible
AI doesn’t replace the lawyer. It reprices the first draft.
Selling hours
- Revenue grows with time spent
- Efficiency shrinks the invoice
- The client pays for the process
AI is a threat to the business model
Selling outcomes and assets
- Revenue grows with judgment and results
- Efficiency widens the margin
- Playbooks become firm assets
Every model improvement works in your favor
One boutique firm in Mexico set its goal as “six people with the capacity of thirty”. That only pays if the pricing changes too.
Line shafts and belts in a Victorian machine shop · photo: Les Chatfield, CC BY 2.0
04Section
01Intelligence is becoming abundant
02It’s worth what you can verify
03What becomes possible
04Why faster lawyers don’t make better firms
05Delegate the work, keep the authority
06The firm that learns
04Why faster lawyers don’t make better firms
95%
of organizations in the study reported no return on their generative AI investment.
MIT NANDA, 2025 · preliminary findings: 52 organizations interviewed, 153 leaders surveyed, 300+ public initiatives. Not a universal rate.
Productive individuals don’t automatically make productive firms.
Central idea · Hebbia
04Why faster lawyers don’t make better firms
We have seen this before: electricity
The bottleneck isn’t the model. It’s the org chart.
Paul A. David, “The Dynamo and the Computer”, American Economic Review, 1990.
04Why faster lawyers don’t make better firms
Faster drafts. Same queue.
Faster lawyersEvery draft waits for the same partner
Redesigned flowEvidence travels with the draft · routine is delegated
Waiting for the partner 0Delivered 0
Waiting for the partner 0Delivered 0
Speed moves the bottleneck. Only redesign removes it.
04Why faster lawyers don’t make better firms
A faster lawyer is not a better firm
The 10× lawyer
- Each lawyer with their own chat
- Faster first drafts, same process
- Prompts and know-how live in one head
Gains stay with the person, and leave with them
The productive firm
- Workflows redesigned around the outcome
- Shared playbooks, context and permissions
- Every correction improves the system
Gains compound across matters and people
The question for leaders: what breaks if this person leaves?
05Section
01Intelligence is becoming abundant
02It’s worth what you can verify
03What becomes possible
04Why faster lawyers don’t make better firms
05Delegate the work, keep the authority
06The firm that learns
05Delegate the work, keep the authority
From operating software to delegating an intention
Today
Five systems. The lawyer is the glue.
With an agent
“Prepare the fee proposal for Alfa Capital: due diligence for the Beta acquisition.”
- Client and matter identifiedCRM
- Scope drafted from our due-diligence playbookprocedure
- Fee calculated with the approved rate cardsame inputs, same fee
- Conflict check still pendingproposal ready, not final
- Client asks for 18% off. My limit is 10%sent to the partner
- Send to the clientneeds a partner’s authorization
05Delegate the work, keep the authority
A task ends.
A responsibility continues.
Delegate executionPrepare a fee proposal
Delegate continuityKeep a matter moving between conversations
- It keeps state: where the matter stands, what is pending
- It shows evidence: every claim points to a source
- It knows its limits: what only you can decide
#project-alfa Matter agent · illustrative, fictional data
MP
Managing partner
How are we doing on Alfa?
A
Matter agent
Since Tuesday:
- The other side returned the SPA markup: 14 changes, 3 material. Summary ready.
- Closing date: the minutes say 14 Nov, the client’s email says 21 Nov. I didn’t choose one.
- Two documents are still missing from the data room. Reminder drafted, not sent.
email · 21 Octminutes · 17 Octdata room index
A
Matter agent
Needs your decision: which closing date, and may I send the reminder?
MP
Managing partner
Where do the 3 material changes come from?
A
Matter agent
Clauses 7.2, 9.1 and 12.4, checked against our SPA playbook. Opening the redline.
05Delegate the work, keep the authority
The application comes apart into pieces an agent can use
The applicationscreens · rules · calculations · data · permissions
Agentcomposes the pieces
KnowledgeWhat a client, a matter, a conflict meanknowledge base
ProceduresHow this firm prepares a fee proposalskills
ToolsValidated calculations, documents, emailMCP · APIs
AuthorityWhat it may do, and for whompermissions
MemoryWhere each matter standsdurable state
The firm’s know-how becomes something an agent can use. Partners still define what “right” means.
05Delegate the work, keep the authority
Separate the intelligence
from the authority
The most trustworthy system won’t be the one built on the model we trust most. It will be the one that lets us trust the model the least. Satya Nadella, CEO of Microsoft · essay, 10 Oct 2026 · paraphrased
Treat the model like a capable new hire: identity, least privilege, logs and approvals.
05Delegate the work, keep the authority
This week · 21 Oct 2026
Found by a model. Fixed by institutions.
“We expect similar findings to increase sharply in the coming months.”OpenAI · Patch the Planet, Oct 2026
Finding it took intelligence. Fixing it took authority, coordination and trust.
05Delegate the work, keep the authority
A plea is not a control
A plea
“Please be careful with confidential information.”
- Lives in the prompt
- Depends on the model behaving
- Cannot be audited
A lock
The agent cannot send, sign or file. A human button does.
- Lives outside the model
- Works even when the model is wrong
- Can be explained to the client
For every agent: what must it never do without someone authorizing it?
06Section
01Intelligence is becoming abundant
02It’s worth what you can verify
03What becomes possible
04Why faster lawyers don’t make better firms
05Delegate the work, keep the authority
06The firm that learns
06The firm that learns
Three lawyers judged the same AI answers. How often did all three agree?
25%of comparisons · 24 legal taskson the three strongest models · 25% across all
If your partners can’t agree on what good looks like, no machine can learn it.
Each square: 1% of comparisons · Harvey & Snorkel, Sep 2026
06The firm that learns
A firm learns when every error changes the system
If the correction stays in one person’s head, the firm didn’t learn. The person did.
This is Toyota’s continuous improvement. What’s new: the system can now draft its own correction, and a human approves it.
06The firm that learns
The partner’s redlines are the firm’s most valuable asset
Works
Agent drafts
with an audit trail
Corrects
Lawyer reviews
accepts, edits, rejects
Learns
Proposes a rule change
every week
Approves
Partner signs off
the playbook updates
Earned autonomy The agent may automate only after consistent agreement with human decisions.
Today those corrections live in tracked changes and email. They evaporate.
Loop based on ABC Legal’s published setup (Anthropic, Aug 2026).
06The firm that learns
Lawyers already know how to do this: it’s called drafting
A wish
“Be careful with indemnities.”
- Can’t be tested
- Means something different to each reader
A rule
“Indemnity caps above 1× fees go to a partner, with the clause cited.”
- Observable and testable
- The same for every reader, human or machine
“A company’s eval suite will be its most valuable resource.” Your precedents and partner reviews are yours.
George Sivulka, Hebbia · Jul 2026
06The firm that learns
The pyramid has a hole at the bottom
If agents do the first draft, how do you train the next partner?
An open question, not a forecast. “Replacing the pyramid” is Luke Drago’s phrase (The Intelligence Curse), Oct 2026.
07Leading through it
Five questions for your next partners’ meeting
01
What service could we offer if this capability were available all the time?
02
Where does work wait today: for data, for review, or for a decision?
03
What may the machine never do without a named person?
04
Which correction do we repeat every week, and who turns it into a rule?
05
How will juniors learn to spot an error in a draft they no longer write?
07Leading through it
The advantage isn’t knowing more.
It’s knowing again.
The blueprint of the AI-native firm is public. Every competitor can read it. What we know about these tools today is worth about ninety days.
What can’t be copied: how fast your firm learns, and who owns that learning.
Intelligence is becoming abundant.
What it’s worth to your firm depends on what you can verify, authorize and learn.
Pick one workflow. Decide what the machine may never do alone. Make every correction count.
Adolfo Rangel Prado
Independent Consultant · Associate Professor, Universidad Panamericana
AAppendix
Sources
- OpenAI · Navier–Stokes, 8 Sep 2026 · mathematical results, 6 Oct 2026
- Ilya Sutskever · on intelligence and human qualities, 2023
- Anthropic · protein binders against scorpion neurotoxins, 20 Oct 2026 (binding only; no antivenom claimed)
- Mark Ajzenstadt · the AI-native services firm, 27 Aug 2026
- OpenAI · Astra for Law, 17 Sep 2026 (vendor’s own 200-question set)
- Harvey · Tenet research preview, 20 Aug 2026
- Anthropic · ABC Legal case, 17 Aug 2026
- MIT NANDA · The GenAI Divide, 2025 (preliminary findings)
- George Sivulka, Hebbia · Productive individuals don’t make productive firms
- Paul A. David · The Dynamo and the Computer, American Economic Review 80(2), 1990
- Satya Nadella · separating intelligence from authority, 10 Oct 2026
- PHP · security advisory, 21 Oct 2026 · OpenAI · Patch the Planet
- Harvey & Snorkel · augmenting human preference in complex domains, 17 Sep 2026
- George Sivulka (Hebbia) via Marc Andreessen · evals as a company’s key resource, Jul 2026
- Luke Drago · on replacing the pyramid, Oct 2026
- Photos: Wikimedia Commons · Steve Jurvetson (CC BY 2.0), Iskra.piotr (CC BY-SA 4.0), Les Chatfield (CC BY 2.0)
Vendor-reported figures are labeled as such. The fee proposal, the matter agent and the review simulation are illustrative, with fictional data.