Main Menu

Alchemy Interactive Framework: From Search Hits to Answers

blank

By John Tredennick

Merlin’s Alchemy Interactive Framework represents a fundamental shift: from passive searching to active collaboration with AI. Instead of reviewing search hits, professionals can now ask questions and get answers.

At the heart of this shift is a simple, intuitive triad:

  • Find – rapidly locate relevant documents in vast data sets.
  • Analyze – let AI models review, classify and synthesize the information that has been found, surfacing insights and patterns.
  • Answer – synthesize across documents to deliver clear, defensible reports.

     

The framework is interactive, meaning users are no longer trapped in a linear “search and review” cycle. Instead, they work alongside AI—refining questions, exploring results, and iterating quickly. It is discovery transformed into intelligence.

 

Step One: Find

The first step addresses the most basic pain point: the inadequacy of keyword search. Boolean logic has been the industry’s crutch for decades, but it cannot capture nuance, context, or evolving legal questions.

Alchemy replaces this with a triad of AI-powered algorithms:

  1. Semantic Search (NLP): Understands intent and meaning, not just literal words.
  2. AI-Enhanced Keyword Search: Expands and maps terms intelligently, reducing the risk of missing synonyms or related concepts.
  3. Machine Learning (CAL heritage): Continuously improves with feedback, surfacing more of what matters and less of what doesn’t.

The result is a search process that is smarter, faster, and more defensible. Lawyers can type a plain-language question like this relating to the BP Oil Spill disaster: 

“I need to analyze all technical testimony regarding the well flow indicators that were present before the Macondo well blowout, including data from monitoring systems, observed anomalies, and then identify all explanations given for why these signals were not adequately recognized or acted upon.”

In seconds, Alchemy’s three types of AI-powered search work together seamlessly finding the 200+ most relevant documents or transcript sections likely to contain information to answer this question. For our part, we haven’t had to expend any efforts on building and testing a complex boolean query. 

What once took days of refining keyword strings can now be accomplished in seconds.

 

Step Two: Analyze

Finding documents is only the beginning. In traditional workflows, this is where armies of reviewers would dig in, plowing through files page by page. That process is slow, costly, and inconsistent.

Alchemy automates this stage through Generative AI-powered document intelligence. Models can:

  • Perform rapid first-pass reviews to determine relevance.
  • Extract key facts, entities, and relationships from documents.
  • Recognize patterns across hundreds of thousands of pages.
  • Generate clear, context-aware summaries that preserve legal nuance.

     

Different AI models are applied strategically. Lightweight models such as Haiku 3.5 or GPT 5 mini can quickly review and assess hundreds of documents for relevance, reading, summarizing and scoring for relevance. What once took days of manual review can now happen in minutes.

For example, here is start of a summary of a highly ranked document prepared by Haiku 3.5: 

Document Type: Court Transcript
Date: 2013-02-28
From: Mark Bly (Witness)
Subject: Deepwater Horizon Oil Spill Trial Testimony

Narrative Overview:
This document section contains testimony from Mark Bly during the Deepwater Horizon oil spill trial, focusing on technical anomalies observed in well monitoring data before the blowout. Bly discusses three specific pressure-related anomalies identified during the investigation.

Key Findings on Well Flow Indicators:

1. First Anomaly [¶176-¶178]:

  • Occurred from 9:01 to 9:08 (2101 to 2108)
  • Drill pipe pressure was increasing when it should have been declining
  • Unexpected pressure increase during fluid movement in the well

     

2. Second Anomaly [¶180-¶182]:

  • Occurred from 9:08 to 9:14 (2108 to 2114)
  • Drill pipe pressure rose by 248 psi after pumps were shut down
  • Pressure should have remained constant when no fluids were being moved

     

3. Monitoring Responsibilities [¶183-¶186]:

  • Drill pipe pressure monitoring was typically the responsibility of:
  • Drillers in the driller’s shack
  • Mud logging personnel
  • Interviews with BP well-site leaders confirmed no additional monitoring parties

     

4. Third Anomaly [¶187-¶188]:

  • Occurred around 9:30 (2130)
  • Approximately 500 psi pressure gain
  • Identified as another instance of unexpected pressure increase

     

Bly emphasized that these anomalies were technically observable and represented potential warning signs before the well control event.

You can review the full set of summaries I created here. 

These lesser models can review, summarize, and rank hundreds of documents in seconds, passing their judgments as training seeds for the CAL algorithm. We use them because they have the intelligence to do the job, particularly in reviewing single documents, but cost as much as 30 times less than the larger models which we use in the next phase of the process. 

Step Three: Answer

The final step is where discovery truly becomes intelligence. Instead of delivering piles of tagged documents or isolated summaries, Alchemy synthesizes across the entire collection to generate answers.

This means:

  • Structured reports with full source citations.
  • Timelines, witness kits, and issue-specific briefs created automatically.
  • Cross-document synthesis that ties facts together into coherent narratives.
  • Interactive refinement—users can ask follow-up questions to explore new angles instantly.

Here is the start of the answer Alchemy prepared to our topic question:

Executive Summary
  • Multiple, independent indicators of influx and loss of well integrity were present before the blowout, beginning with an anomalous negative pressure test and followed by progressively clearer surface signals during displacement. These included drill pipe pressure increases with pumps off, flow-out exceeding flow-in, pit gains, choke/kill line pressure anomalies, and visible surface flow. Real-time monitoring capability existed through two parallel systems (Transocean Hitec and Sperry Sun), but visibility was degraded by overboard diversion, sensor placement, lack of schedules, and simultaneous operations.

  • The negative pressure test anomalies were repeatedly misinterpreted and ultimately accepted as a pass, creating a false mental model that the well was secure and contributing to reduced vigilance during displacement. Operational and organizational factors (bypassed sensors, lack of shared data and schedules, distractions, evolving procedures, and expectation bias) compounded delayed recognition and inaction.

  • Expert analyses conclude the indicators, if properly recognized and acted upon (manual flow check and immediate shut-in), could have prevented or mitigated the incident. Responsibility assessments highlight BP and Transocean decision-making, with specific testimony that the mudlogger lacked key data access during the critical period.

You can download and read the entire answer here.  You can also watch a video of me running the topic here. 

Creating a Closing Argument: This report is only the beginning. I asked Alchemy to review the entire BP Oil Spill trial record (50+ days of transcripts, exhibits, reports, and depositions) and create a comprehensive closing argument on behalf of BP. In less than 90 minutes, we were able to generate a comprehensive 60-page closing argument for BP’s defense, integrating testimony and documentary evidence alike. You can see the outline Alchemy created along with the full text of the closing argument I created for BP here. 

How about Medical Records? 

As another example, consider medical records. In complex litigation, tens of thousands of pages often must be analyzed to establish treatments, diagnoses, or causation. Traditionally, paralegals and nurses would spend weeks building chronologies. 

With Alchemy, professionals simply load the records into the system. In seconds, the platform can answer detailed questions about medical history, providing comprehensive, defensible answers. What once drained budgets and delayed strategy now delivers immediate clarity.

For example, imagine coming up with effective keywords to answer this question:

How was Patient A’s atrial fibrillation managed during her hospitalization, including rate control, anticoagulation, and monitoring for complications? 

Rather than struggle with keywords, we just put the question to Alchemy and let it provide this answer: 

Patient A’s Atrial Fibrillation Management During Hospitalization

Patient Demographics and Clinical Context

Patient A was an 86-87 year old female with a documented history of permanent atrial fibrillation who was hospitalized from September 12-23, 2022 at Facility A Medical Center following a minor motor vehicle accident TAM_00000606 §44 ¶676, TAM_00000606 §44 ¶689, TAM_00000606 §38 ¶571, TAM_00000606 §81. The patient had a CHADS2 score of 3, indicating higher stroke risk due to age (>74), hypertension, and diabetes TAM_00000605 §8 ¶99, TAM_00000605 §6 ¶72-¶73.

Anticoagulation Management

Primary Anticoagulation Strategy

Patient A was maintained on Eliquis (apixaban) 2.5 mg twice daily throughout her hospitalization TAM_00000606 §77 ¶1444, TAM_00000606 §73 ¶1363-¶1364, TAM_00000606 §45 ¶696, TAM_00000606 §60 ¶1069, TAM_00000606 §81 ¶1538, TAM_00000606 §42 ¶637, TAM_00000606 §38 ¶574-¶575. This medication served a dual purpose, providing both anticoagulation for atrial fibrillation and DVT prophylaxis TAM_00000606 §73 ¶1363, TAM_00000606 §40 ¶598, TAM_00000606 §60 ¶1090.

Continuity of Care

The medical team ensured continuous anticoagulation throughout hospitalization TAM_00000606 §44 ¶685. The patient’s home medication was continued during hospitalization TAM_00000606 §42 ¶637, TAM_00000606 §38 ¶574-¶575, and she was consistently documented as “adequately anticoagulated” TAM_00000606 §70 ¶1296, TAM_00000606 §76 ¶1412, TAM_00000606 §67 ¶1214.

You can download and read the entire answer here.  You can also watch a video of me running the topic here. 

As you will quickly see, the shift is dramatic: from searching for needles in haystacks to receiving detailed, defensible answers backed by the full record.

 

Why It Matters: A Workflow Revolution

The Alchemy Framework doesn’t just speed things up. It re-engineers the very nature of legal work.

  • Efficiency: Review hours cut by 90%, hosting costs reduced by up to 60%.
  • Quality: Consistent analysis, fewer missed documents, stronger pattern recognition.
  • Strategy: Faster case assessments, deeper fact development, better client counseling.

     

Instead of spending time sorting documents, lawyers can spend time being lawyers—exercising judgment, developing strategy, and advising clients.

This is why we call Alchemy the first Document Intelligence Platform. It’s not about search—it’s about insight.

 

Practical Applications Across the Legal Spectrum

The Alchemy Interactive Framework has already proven its value in a wide range of contexts:

  • Internal Investigations: Rapidly uncover misconduct, compliance failures, or fraud.
  • Litigation Discovery: Move from weeks of manual review to near-instant fact development and strategy.
  • Regulatory Compliance: Respond quickly to government requests with defensible reports ready in hours.
  • Due Diligence: Analyze contracts and risks across thousands of pages in record time during M&A.

     

One global client used Alchemy to process more than 100,000 documents during a regulatory investigation. The platform delivered evidence-backed answers in weeks, not months or years—fully sourced and ready for quality control. That speed fundamentally changed their approach to strategy and negotiation.

 

Built for the Cloud Era

Underneath the framework lies an infrastructure purpose-built for today’s demands:

  • Single-Tenant Security: Every client site operates in its own private cloud environment, eliminating commingling and cross-client risks.
  • Cloud Utility Pricing: Clients pay by the minute, not the month—full rate when sites are active, 70% less when idle. That flexibility cuts hosting costs by up to 60%.
  • Sustainability: Powering down idle servers reduces energy use and carbon footprint, supporting corporate ESG goals.

     

The framework is not only more intelligent—it’s more secure, economical, and sustainable.

 

The Future of Discovery

The Alchemy Interactive Framework points toward the next era of legal practice. As AI advances, discovery will move from reactive searching toward predictive and prescriptive intelligence—helping lawyers not only find facts but also anticipate outcomes and recommend strategies.

This transition demands new skills, workflows, and expectations. Professionals who embrace AI-enhanced tools will deliver faster insights, stronger cases, and greater value. Those who cling to outdated methods risk falling behind.

 

From Information to Intelligence

We never figured out how to turn lead into gold—despite our alchemy namesake. But with cloud technology and a suite of AI tools, we built a platform that turns information into intelligence.

From Find to Analyze to Answer, Alchemy redefines what is possible: turning information into intelligence and questions into answers—in seconds rather than weeks.

 

Ask Questions > Get Answers > In Seconds.
blank

About the Author

John Tredennick (jt@merlin.tech) is the CEO and Founder of Merlin Search  Technologies, a software company leveraging generative AI and cloud technologies to make investigation and discovery workflow faster, easier, and less expensive. Prior to founding Merlin, Tredennick had a distinguished career as a trial lawyer and litigation partner at a national law firm. 

With his expertise in legal technology, he founded Catalyst in 2000, an international ediscovery technology company that was acquired in 2019 by a large public company. Tredennick regularly speaks and writes on legal technology and AI topics, and has authored eight books and dozens of articles. He has also served as Chair of the ABA’s Law Practice Management Section.

Scroll to Top