A Working
Guide to Today's
LLM Tools.
How INFLXD thinks about AI.
At INFLXD, AI is treated as amplified intelligence. Every role pairs strong human judgment with the right tool. AI is not a replacement layer or a productivity metric tracked on a sidebar.
The bar is whether a tool produces something fundamentally better. Not just faster, not just cheaper. If a workflow can be redesigned around AI to produce a different outcome, that is where the energy goes.
This guide reflects what INFLXD has vetted, tested, and continues to use across the company. The recommendations are filtered. The tools listed here are the ones that have earned a place in actual daily work.
AI and LLM are not the same thing
AI is the umbrella term covering any system that performs tasks once thought to require human intelligence, from spam filters to self-driving cars. An LLM, or Large Language Model, is a specific class of AI built to read and generate text. Every tool in this guide is an LLM or an application built on top of one. This guide uses LLM as the precise term going forward.
Where these tools came from
The technical foundation arrived in 2017 with the publication of the Transformer architecture by researchers at Google. The cultural arrival came in November 2022 with the public launch of ChatGPT. Read a brief primer on LLMs.
What this guide is not
This is not a comprehensive AI survey. It is not a how-to-prompt manual. It is not a vendor catalogue. It is a curated set of options, the tools INFLXD has vetted and uses, alongside the workflows that make them useful in practice.
Tools at a glance.
The matrix below maps each tool to the use cases it does well. Tools marked with the pill are the ones INFLXD uses internally and stands behind.
| Tool | Thinking partner | Writing & polishing | Real-time search | Deep research reports | Synthesize your documents | Meeting & voice capture | Works inside your tools |
|---|---|---|---|---|---|---|---|
| Claude Chat* | |||||||
| Claude Cowork* | |||||||
| Claude Code* | |||||||
| Gemini (chat) | |||||||
| Gemini in Workspace | |||||||
| NotebookLM* | |||||||
| Deep Research | |||||||
| ChatGPT | |||||||
| Perplexity | |||||||
| Wispr Flow* | |||||||
| Granola* |
The toolkit: Claude and Google AI.
Claude
Anthropic's family of LLM products, used at INFLXD as the default for thinking, writing, building, and operating on files. Three surfaces, each for a different mode of work.
Claude Chat *
A thinking partner for strategy, decisions, and high-stakes writing. Claude produces first drafts that are closer to publishable than the alternatives, particularly when matching a specific voice or tone.
Best for: pre-decision sparring, drafting investor or board communications, turning lengthy reports into one-page briefings.
- Set up Projects for each recurring workstream. Upload relevant documents once and every conversation inside the project carries that context.
- Connect calendar, Gmail, and Drive through Connectors so questions like "what is on my plate tomorrow" are answered against real schedule data.
- Tell Claude how it should push back. Adding "challenge my assumptions and flag weak logic directly" turns the default polite assistant into a useful sparring partner.
Claude Cowork *
An operator that works on actual files and applications while the user does something else. The difference between asking an LLM to advise on work and asking it to do the work.
Best for: morning briefings compiled from a folder of inputs, recurring weekly reports, processing customer interview transcripts into themed summaries.
- Set up a writing style and preferences file. Drop in samples of past memos, decks, or emails so Cowork can match the voice on every draft from then on.
- Create one project per recurring deliverable, each with its own folder of templates, examples, and rules.
- Add "start with a clarifying question before executing" to standard prompts. A short alignment up front prevents a full redo at the end.
Claude Code *
Anthropic's agentic coding system. A version of Claude that can read, write, and edit files on a computer the way a junior engineer would, driven by plain-English instructions. The user describes what they want ("build a small dashboard that pulls weekly metrics into one view") and Claude Code writes the code, tests it, fixes its own mistakes, and hands back something that runs. It is the most capable AI coding tool on the market in 2026.
Best for: building internal tools without engineering time, reading folders of local files for synthesis, automating repeatable file-based workflows.
- Start with a throwaway tool, not a real product. Ask it to build a tiny webpage that reformats a messy CSV. Low stakes, fast learning.
- Use it as a document synthesizer first. Point it at a directory of meeting notes or PDFs and ask for themes, gaps, or a one-pager summary.
- Pair it with a chief of staff or executive assistant for setup. Installation is a twenty-minute task best delegated, then the executive stays in plain-English mode.
A few skills to get you started
Skills are reusable workflows installed once and available in every Claude conversation that follows. INFLXD's library currently covers end-of-day reports, glossary building, style and tone matching, writing cleanup, data summaries, and document cleanup.
Google AI
Google's AI surface is broad. The four offerings below are the most relevant ones.
Gemini (chatbot)
Google's general-purpose AI assistant. Strongest when the user already lives inside the Google ecosystem and wants multimedia handling. Gemini natively processes video and audio, with context windows up to one million tokens.
Best for: synthesizing mixed-media inputs (a meeting recording, a deck, a whiteboard photo) in a single pass.
- Open Saved Info in Gemini settings and add a short profile (role, team, communication style, recurring projects) so every response reflects the user's context without re-explaining each time.
- Create a Gem from the Gem manager for each repeatable task (board prep, weekly review drafting, hiring rubric scoring) with custom instructions and reference files, then pin it for one-click reuse.
- Turn on the Google Workspace connection under Apps so Gemini can pull directly from Gmail, Drive, Docs, and Calendar when answering questions about real work.
Gemini in Workspace
Gemini built directly into Docs, Sheets, Gmail, Slides, Meet, and Drive. Included at no additional cost in Google Workspace Business and Enterprise plans. For executives whose daily work already lives in Google, this is the most seamless AI experience available.
Best for: drafting documents from connected files, populating spreadsheets from prompts, finding answers across email threads, agentic scheduling.
- Confirm Gemini is enabled at the organization level. Most Workspace Business and Enterprise plans now include it.
- Use the side panel in Docs to draft from existing files in the user's Drive rather than starting from a blank page.
- Try the Gemini side panel in Gmail to summarize and synthesize across email threads rather than reading each one individually.
NotebookLM *
A source-grounded research and synthesis tool. NotebookLM answers only from documents the user uploads. It cannot draw on outside knowledge, which dramatically reduces hallucination by grounding every answer in those sources. The standout feature for executives is Audio Overviews, which converts uploaded documents into listenable podcast-style briefings in Brief, Critique, or Debate format.
Best for: turning a fifty-page report into a fifteen-minute audio briefing, verifying claims against source documents before publishing, pre-meeting prep on a stack of materials.
- Upload all relevant source materials before asking any question. The tool's value comes from its grounding in those sources.
- Try the Audio Overview feature on a long document. It is the fastest way to absorb dense material before a meeting.
- Use Interactive mode to converse with the AI hosts during the audio briefing, asking them to dig into specific sections.
Deep Research
An autonomous research agent that conducts multi-step investigations across the web and connected private data, producing cited multi-page reports with charts and infographics. Built for the work an analyst might spend several hours on. Visual reports (charts, diagrams, interactive simulations) are available to Google AI Ultra subscribers.
Best for: competitive intelligence, market research, technical due diligence, internal knowledge synthesis.
The toolkit: ChatGPT and specialist tools.
ChatGPT
OpenAI's flagship product and the most versatile single LLM tool available in 2026. ChatGPT is the strongest mobile and voice experience and a platform leader in autonomous task execution through ChatGPT Agent and Workspace Agents.
Best for: general-purpose work across a wide range of tasks, autonomous task execution that runs in the background, deep research with source citations.
- Use Projects to encapsulate repeated workflows with stable inputs. A research assistant for one domain, a code reviewer with team style guides built in.
- Try Workspace Agents for scheduled, team-shared autonomous tasks. A Friday report agent, an inbound prospect scoring agent.
- Use the voice mode on mobile for thinking-out-loud sessions during travel or between meetings.
Perplexity
An answer engine built on live web search with citations. Useful for verification and real-time research. Perplexity also offers a Deep Research mode for heavier multi-source work.
Best for: verifying specific facts before publishing or presenting, grabbing real-time information quickly, fact-checking AI-generated content from other tools.
- Use Perplexity as the verification step after drafting in Claude or ChatGPT.
- Apply search operators (for example, site:reddit.com) to narrow results to real user experiences instead of marketing content.
- Bookmark frequently-used queries for daily reference.
Wispr Flow *
A dictation tool that turns speech into clean formatted text in any application. Email, Slack, browser, Claude itself, mobile. Wispr Flow is the lowest-friction productivity tool in this guide, and reaches roughly four times the speed of typing for most users.
Best for: dictating long emails or documents, giving an LLM minutes of spoken context that would take five times longer to type, voice-first drafting when the idea is half-formed.
- Add a custom vocabulary on day one. Company name, product names, team members, recurring jargon. Accuracy jumps noticeably for frequently-used terms.
- Use Command Mode to edit existing text, not just dictate new text. Highlight anything, hit the command shortcut, and say "make this more concise."
- Create snippets for repeated phrases. Saying "my booking" can expand to the full Calendly link automatically.
Granola *
A meeting transcription tool that captures device audio directly from the user's system. No bot joins the call, no recording announcement. Granola transcribes the conversation and produces clean structured notes seconds after the meeting ends.
Best for: staying present in meetings without taking notes, prepping for any meeting by querying past meeting history, searching across a quarter of meetings for patterns.
- Build custom Recipes for recurring meeting types (board meetings, customer calls, one-on-ones). Configure once, get the right format every time.
- Jot triggers during the meeting. Type "DECISION" or "ACTION" at key moments. Granola builds the summary around those triggers.
- Connect Granola output to Claude for cross-meeting analysis.
Other tools worth knowing
ElevenLabs
Voice cloning, AI-generated audio, and conversational voice agents. Useful for branded podcasts, audio content production, or localized voice versions of written material. Not a daily-use tool for most executives, but the leading product in its category.
Other coding tools
Cursor is the leading alternative for engineering teams that prefer a more traditional editor experience. GitHub Copilot is Microsoft's older offering, widely deployed inside large enterprises. Replit, Lovable, v0, and Bolt are browser-based tools for non-developers who want to build a webpage or small app without installing anything.
INFLXD's pick remains Claude Code.
Sales and outbound tools
The tools below sit outside the catch-all LLM toolkit. Each one is purpose-built for a specific sales, outreach, or acquisition job. Worth knowing if any of these workflows are part of the reader's role.
| Tool | Prospect data | Outbound email & LinkedIn | Pipeline & CRM | Landing pages & paid acquisition |
|---|---|---|---|---|
| Apollo | ||||
| Lemlist | ||||
| Waalaxy | ||||
| Pipedrive | ||||
| Lovable | ||||
| Google Ads | ||||
| LinkedIn Ads |
Apollo
An AI-powered B2B sales platform that combines a contact and company database with outbound sequencing in one system.
Best for: prospecting into a defined ICP and running multi-touch outbound to verified contacts without stitching together separate data and engagement tools.
Lemlist
A sales engagement platform for running personalized cold outreach across email, LinkedIn, calls, WhatsApp, and SMS from a single campaign.
Best for: sending personalized cold email sequences at scale with deliverability controls (via its lemwarm warm-up tool) and multichannel follow-ups.
Waalaxy
A LinkedIn automation tool that sends connection requests, messages, and email follow-ups inside multi-step campaigns from a Chrome extension and web app.
Best for: running automated LinkedIn prospecting sequences (with optional email steps) without manual sending or technical setup.
Pipedrive
A sales CRM built around a visual, kanban-style pipeline and activity-based selling.
Best for: tracking deals through a defined sales process and forecasting revenue against pipeline coverage.
Lovable
An AI app builder that turns natural-language prompts into working full-stack web apps and landing pages, with editable code, hosting, and integrations included.
Best for: shipping a working prototype, landing page, or internal tool quickly by describing it in plain language instead of building from scratch.
Google Ads
Google's paid advertising platform for running search, display, shopping, YouTube, and Performance Max campaigns across Google's properties and partner network.
Best for: capturing existing demand at the moment of intent, especially through keyword-targeted search ads tied to commercial queries.
LinkedIn Ads
LinkedIn's self-serve advertising platform (managed through Campaign Manager) offering Sponsored Content, Message Ads, Dynamic Ads, and Text Ads targeted by professional attributes.
Best for: reaching specific B2B audiences by job title, company, seniority, and industry when the buyer profile is more targetable than the search intent.
An INFLXD product worth highlighting
Expert Network News is an INFLXD-built AI product that aggregates and analyzes daily industry news. It powers a weekly newsletter and podcast and is part of INFLXD's inbound marketing flywheel rather than a tool the reader installs.
How the tools work together.
Different tools handle different phases of work. The flowchart below matches the phase to the recommended tool.
Sample workflows.
Five short, replicable workflows built around real situations. Each one names the tools, the sequence, and the output.
01 · Post-meeting synthesis from a long call
Granola captures and transcribes the meeting in the background. After the call, the transcript is handed to Claude with a prompt like "produce a one-page summary with decisions made, action items by owner, and open questions." The output is a clean briefing ready to share within minutes of the meeting ending.
02 · Briefing prep from a folder of raw inputs
Drop a folder of customer interview transcripts, board notes, or market research into a Cowork project. Ask for the top themes, supporting quotes, and a draft summary. The role shifts from processing material to editing the draft and adding judgment.
03 · Pre-decision sparring before a major call
Share the decision (an acquisition, a hire, a major commitment) with Claude and ask it to argue against the decision from several angles before stress-testing the assumptions. The exercise surfaces blind spots before they become expensive.
04 · Competitive intelligence on a market
Deep Research produces a cited, multi-page report on a competitor or a market segment. Perplexity is then used as the verification scalpel. Checking specific figures, claims, and dates before any of the output gets shared with stakeholders or used in a decision.
05 · Long-document compression for absorption on the move
A fifty-page report, a board pack, or a regulatory filing is uploaded to NotebookLM. The Audio Overview feature converts it into a fifteen-to-twenty-minute listenable briefing.
The workflows above are not prescriptive. They are examples of how the tools combine in practice. Pick the situation that resembles work already on the calendar, and try that sequence first.