How we prepare firms for AI

From scattered processes to documented workflows ready for AI

The Problem

Their client onboarding process existed in 5 different people's heads. New hires took 6 months to learn it. They wanted AI to help with onboarding but had nothing documented for AI to learn from.

Clarity Package: What We Did

We spent 4 weeks interviewing their team and documenting how onboarding actually worked.

• Mapped the entire onboarding process step-by-step across 127 individual steps
• Identified 23 decision points where someone makes a judgment call
• Documented the criteria they use to make each decision
• Found their templates scattered across 3 systems and organized them in one knowledge base
• Created decision trees showing when to use which template
• Wrote down the tribal knowledge that only senior staff knew

The Result

For the first time, they could see their entire process written down. They realized 40% of their steps were redundant. Before implementing any AI, they simplified their process. Now AI had something clear to learn from.

Implementation: What We Built

Because the process was documented, we could build AI that actually understood their workflow.

• Built AI assistant that guides staff through onboarding using the documented decision trees
• Connected AI to their organized knowledge base so it pulls the right templates
• AI suggests next steps based on the 127-step workflow we mapped
• Trained their team using the same documentation we created in Clarity phase

Partnership: Ongoing Refinement

Six months later, they added a new service offering. Because their processes were already documented, adding it to the AI system took days instead of months.

Journey: Clarity Package → Implementation → Partnership

Organizing 15 years of case files so AI could learn from past work

The Problem

They had 15 years of case files with no consistent naming, no tagging system, and no way to find past similar cases. They wanted AI to suggest strategies based on past wins but their data was chaos.

Strategy Package: What We Did

We spent 8 weeks creating a data structure and classification system.

• Analyzed their document structure and identified 12 types of documents they create
• Created a tagging taxonomy based on case type, outcome, jurisdiction, and complexity
• Designed a naming convention that made files searchable
• Built a classification guide showing staff how to tag new cases
• Created a knowledge base structure where AI could access precedents
• Mapped which documents need to stay confidential vs which can train AI

The Foundation Work

Before implementing AI, their team spent 3 months reclassifying their most important 200 cases using the new system. This wasn't AI work. This was organizing their knowledge so AI could use it.

Implementation: What We Built

Now that their data was organized, AI could actually learn from it.

• Built AI that searches their reclassified case files based on current case parameters
• AI suggests relevant past cases using the taxonomy we created
• Pulls specific clauses and arguments from similar past wins
• Respects the confidentiality rules we mapped during strategy phase

The Difference

Without the 8 weeks of strategy work organizing their data, the AI would have been searching chaos. Now it searches a structured knowledge base.

Journey: Strategy & Roadmap → Implementation

They had everything documented. We built AI that could read it.

The Situation

Unlike most firms, they already had documented processes, organized data, and clear workflows. They knew exactly what they wanted AI to do. They just needed someone to build it.

What We Built

Because their foundation was solid, we could move straight to implementation.

• Built AI that generates first drafts of client reports using their methodology documentation
• Trained AI on their writing style guide and quality standards
• Connected AI to their research database so it pulls relevant data
• Set up review workflows where senior consultants approve AI output
• Created templates that AI fills in based on project parameters

Why It Worked Fast

Most implementation projects take 12-16 weeks. Theirs took 8. The difference was their documentation. AI had clear rules to follow, clear data to access, and clear quality standards to meet.

The Lesson

AI is only as good as the foundation you give it. They spent years building that foundation before we arrived. That's why implementation was smooth.

Journey: Implementation (foundations already in place)

Discovered they weren't ready for AI. Saved them from wasting money.

What They Thought They Needed

They wanted AI to automate contract review. They were ready to spend significant money on AI tools.

Clarity Package: What We Found

In 4 weeks of assessment, we uncovered fundamental issues.

• Their contract templates weren't standardized across the 3 partners
• Each partner had their own review criteria that contradicted the others
• No documented checklist of what makes a contract acceptable
• Their "process" was actually 3 different processes
• No central repository of past contracts organized by type

Our Recommendation

Don't implement AI yet. First, get your partners to agree on one process. Standardize your templates. Document your review criteria. Then come back.

What They Did

They spent 6 months getting their house in order. They didn't work with us during that time. They just fixed what we identified.

The Value

We saved them from implementing AI on top of chaos. If they had bought AI tools without fixing their foundation, it would have automated 3 conflicting processes. That's worse than no AI at all.

Journey: Clarity Package (stopped here, foundations need work)

AI systems decay without maintenance. We keep theirs current.

After Implementation

We built them an AI system for tax return preparation. It worked great for 4 months. Then it started giving outdated advice.

The Problem

Tax regulations change constantly. Their AI was trained on rules from implementation. Six months later, those rules had changed.

Partnership: Ongoing Updates

Now we update their system monthly.

• Monitor for changes in tax regulations
• Update the AI's knowledge base when rules change
• Retrain decision logic based on new case law
• Add new document types as their practice expands
• Refine prompts based on what's working and what's not
• Upgrade to new AI models when they're better for tax work

The Reality

AI isn't set-and-forget. The knowledge bases need updating. The models improve. The regulations change. Without partnership, their AI would be giving dangerous advice based on old rules.

What Partnership Includes

Monthly calls to review what's working. Quarterly updates to knowledge bases. Immediate fixes when regulations change. Training for new staff. Documentation updates as processes evolve.

Journey: Implementation → Partnership

They thought they needed AI. They actually needed standardized processes.

The Request

They wanted AI to speed up their proposal generation process. Currently taking 8 hours per proposal.

Clarity Package: What We Found

Their proposal process wasn't slow because of lack of AI. It was slow because everyone did it differently.

• Each team member had their own proposal template
• No standard structure for what information proposals need
• Team spent hours searching for past proposals because they weren't organized
• Pricing calculations done manually with different formulas
• Past case studies scattered across individual computers

The Real Problem

AI can't speed up a process that has no standard. We showed them that 6 of those 8 hours were wasted looking for information and reconciling different approaches.

What They Actually Needed

One standard proposal template. One central repository of past proposals organized by industry. One pricing calculator. One library of approved case studies.

The Outcome

They implemented those four things without AI. Proposals now take 3 hours instead of 8. They saved more time by fixing their process than AI would have saved automating their mess.

The Clarity Value

We prevented them from spending money on AI to automate chaos. Sometimes the answer isn't AI. Sometimes it's just good process design.

Journey: Clarity Package (AI not needed yet)