CounselExpress

Day 10 of 30 — Cogle GTM 1/3

CounselExpress is privilege-preserving litigation support for personal injury attorneys. Live today at counselexpress.app. It’s product 1 of 3 in the Cogle GTM cluster announced yesterday on Day 9 — the AI-native services firm Jim Hasty and I now operate as general partners.

The original working name on the bench was CasePulse. The launch-day name is CounselExpress — the rename happened over the back half of week 1 once the four-analysis structure stabilized and “case pulse” stopped describing what the product actually does. The product reads case files and produces counsel-ready analyses; the express in the name carries the velocity claim. The brand maps to what the product is.

The site is a single landing page at counselexpress.app. The desktop app itself is the product — Electron, files-stay-local, Claude Sonnet over AWS Bedrock. Today’s launch is the public surface and the early-access waitlist; cohort 1 opens once the full Cogle GTM cluster is live (Day 12, May 12).

The ruling that frames the launch

In United States v. Heppner (S.D.N.Y., February 2026), a federal court ruled that information a defendant input to a consumer generative AI platform on his own initiative is not protected by attorney-client privilege or work-product doctrine. The court found that all recognized privileges require “a trusting human relationship,” that no such relationship exists between an AI user and a public platform, and that inputs to that platform are shared with a third party — which waives the privilege the same way handing the document to opposing counsel would. “Anthropic’s Claude is not a licensed attorney” is the load-bearing line in the opinion.

The court drew one explicit line: enterprise-tier deployments with contractual confidentiality protections (Claude over AWS Bedrock, Claude commercial / government plans, ChatGPT Enterprise) are different. Those agreements exclude inputs from training and bind the provider to confidentiality terms. The consumer-tier subscriptions don’t.

For solo and small-firm PI attorneys, Heppner is operationally enormous. The structural information asymmetry between PI plaintiffs’ counsel (small headcount, no paralegal team) and defense counsel (insurance-funded paralegal teams) was already in defense’s favor; the obvious AI shortcut to close the gap just got ruled out for privileged work. CounselExpress is built specifically for the path Heppner left open.

What it is

CounselExpress is a desktop application — not a SaaS — that points at a folder of case files and produces four structured outputs:

• Medical Timeline. Chronological treatment history across every provider, costs matched against billing records, treatment-gap detection. The kind of timeline defense counsel will use to argue causation gaps — built first, by plaintiffs’ counsel, before defense raises the argument.
• Case Dashboard. Statute-of-limitations countdown by jurisdiction and cause of action. Service-of-process status. Discovery deadlines. The running scoreboard of what the firm owes the court and what the other side owes the firm.
• Police Report Analysis. Fault and liability extraction, witness and evidence inventory, plus the contradictions inside the responding officer’s narrative the report itself does not surface — cross-examination prep, on first read.
• Discovery Analysis. Per-request deficiency review under the relevant rules of civil procedure, with O.C.G.A. § 9-11-37 awareness for Georgia practice. Includes a draft 6.04 letter / motion-to-compel skeleton that’s edit-ready, not chat-transcript output.

The privilege-preserving stack

Three architectural choices put CounselExpress on the right side of Heppner:

1. Inference runs on AWS Bedrock, not the consumer Claude API. Same model. Different legal posture. Bedrock inputs are not used for training and are bound by AWS’s enterprise data-isolation terms — the contractual-confidentiality envelope the Heppner opinion identified as the privilege-preserving path. This is the choice that makes CounselExpress usable for a privileged matter; without it, an attorney clicking “run analysis” would be doing on Saturday morning exactly what Heppner held waives privilege.
2. Case files never leave the laptop. Electron desktop application. PDFs sit in a folder on the attorney’s machine. pdf-parse does extraction on-device. better-sqlite3 holds the structured analyses locally. Only the specific extracted text needed for the requested analysis ever crosses the wire to Bedrock — and only when the attorney triggers it.
3. Civil-procedure aware output, not chat-style. The Discovery Analysis is structured around the rules attorneys actually cite. Output is cite-checkable and edit-ready. The attorney keeps the judgment layer; CounselExpress takes the reading and structuring layer.

Encrypted credentials via electron-store. Sandbox is disabled (required for the native SQLite module) but context isolation is on, so the renderer has no direct Node.js access.

The thesis behind the launch

CounselExpress is the first product in a CogleGroup line built explicitly around line professionals doing skilled, repetitive, emotionally draining work inside regulated processes. The PI attorney reading 3,000 pages of medical records on a Saturday is doing the kind of work nobody got into the law to do. The chamber-of-commerce coordinator hand-emailing 800 members about an event is doing the kind of work nobody got into community development to do. The county planner reviewing the same site-plan checklist for the forty-third time this month is doing the kind of work nobody got into urban planning to do.

That is the customer the Cogle line is built for. Not “AI replaces the line professional.” AI takes the part of the job nobody got into the work to do, and gives it back as time. CounselExpress runs that play in litigation this week. Day 11’s ChamberAdvance runs it for chamber engagement coordinators. Day 12’s PlanCheckers runs it for county planners. By Wednesday night the cluster proves the shape generalizes — same Velocity Process, same Bedrock-enterprise discipline, three deliberately distinct line professionals.

This is also why the regulated-legal opener mattered. Heppner is the moment that makes the privilege-preserving stack a category, not a feature. CounselExpress is what that category looks like for plaintiffs’ PI practice.

Pricing & business model (working)

Per-seat desktop license, $149/month or $1,490/year per attorney. Volume pricing for 5+ seats. Cohort 1 (opening Day 12, May 12) is free for 30 days in exchange for two 30-minute feedback calls — a research mechanism as much as a launch mechanism. The cohort is intentionally biased toward solo and small-firm PI practice in Georgia for the first wave so the discovery-rules code path can ship deeper Georgia-specific behavior before generalizing.

The product does not bill by token. The attorney sees per-analysis Bedrock cost in the UI; the firm pays the per-seat subscription. Sonnet-default keeps per-analysis cost in the cents-to-low-dollars range for a typical case folder.

The Velocity Process notes

What Claude Code handled: the entire counselexpress.app Astro 5 site (BaseLayout with the SoftwareApplication + CogleGroup Organization + WebSite JSON-LD graph, Nav, Footer with the Cogle cluster references and the line-employee thesis breadcrumb, single-page landing with hero / Heppner expansion / problem / four-analyses / privilege-preserving stack / “why we built this” thesis section / Cogle GTM / waitlist sections, llms.txt and robots.txt explicitly welcoming Claude-Web / GPTBot / Perplexity / anthropic-ai); the rename sweep across the silverbackcto.com Day 9 launch post (CouncilExpress → CounselExpress in the launch announcement, CasePulse → CounselExpress in the “what’s next” section and the closing line); the .law → .app domain sweep when the day-of decision moved off the .law TLD; the waitlist form wired to the existing GHL webhook with a counselexpress-waitlist tag (no new GHL sub-account); the AWS S3 + CloudFront + Route53 deploy plan as a DEPLOY.md in the repo so the launch-day deploy is one npm run build && aws s3 sync && cloudfront create-invalidation command.

What required human judgement: the rename itself (CasePulse → CounselExpress — the four-analysis structure made CounselExpress the right brand once the product crystallized); the choice to lead the launch with the Heppner ruling rather than with feature copy — Heppner is the structural reason any PI attorney needs CounselExpress today, and feature copy is the wrong frame in front of a ruling that operationally bans the consumer alternative; the line-professional thesis as the unifying frame across the Cogle cluster (litigation paralegals, chamber coordinators, county planners — same shape, different vertical), which is the answer to “is CounselExpress a one-off or the start of a category”; the .app TLD choice for the marketing site; the Bedrock-only architectural posture (the simpler thing was to wire the consumer Claude API and call it good — Heppner makes the simpler thing unusable for the customer, so we ate the Bedrock setup time); the early-access cohort scoping (Georgia-first, solo / small-firm PI, 30 days free in exchange for two feedback calls — real research mechanism, not faux exclusivity).

What broke: brand-name continuity. Yesterday’s Day 9 launch post went out referring to today’s launch in two different names — CouncilExpress in one place (one-letter typo) and CasePulse in another (the working name the rename had already superseded). Both got swept this morning before the Day 10 site went up, plus a forward-looking sweep across the GTM calendar, the Sequoia copy, the Week 1 recap email, and the Substack halftime piece so Day 11 and Day 12 don’t inherit the same problem. The lesson: when a venture renames mid-week, the rename has to land in every public document the same morning the rename closes — a 24-hour gap is enough for the next day’s launch post to inherit two stale names. The fix is a pre-launch grep across the silverbackcto.com content collection plus the GTM/content folders the morning of each launch, not just the launch-day file.

What I’d do differently: lock the rename two days ahead of the launch instead of the morning of. The product crystallized on Day 7; the rename should have closed Day 8. Same lesson on architectural posture: Bedrock vs. consumer API was a decision that should have been made on Day 7 too, not relitigated this morning when the Heppner significance became impossible to ignore.

What’s next this week

• Day 11 (Mon May 11): ChamberAdvance — Cogle GTM 2/3, chamber-of-commerce engagement at AI speed, already in revenue with a metro-Atlanta chamber. Same line-professional shape (chamber engagement coordinators).
• Day 12 (Tue May 12): PlanCheckers — Cogle GTM 3/3, planning-permit software running today at foco.plancheckers.com for a major Atlanta-area county. Same line-professional shape (county planners). The cluster closes here.
• Day 12 evening: CounselExpress cohort 1 opens. First wave is Georgia PI attorneys in solo and small firms; expansion to other jurisdictions begins with cohort 2 in early June.
• Day 13 (Wed May 13): PodToSite — the Discoverability arc opens. Cogle cluster steps back; Websites-for-LLMs theme takes the week.

Want to talk

If you’re a personal injury attorney in solo or small-firm practice — especially in Georgia for the first wave — get on the early-access list at counselexpress.app. 30 days free, two 30-minute feedback calls, real product. Cohort 1 opens Day 12 (May 12).

If you’re an attorney in any regulated practice area trying to figure out what Heppner means for AI use in your firm — the privilege-preserving-stack section on counselexpress.app is a public reference for the architectural shape. Bedrock-routed inference + on-machine processing is the working answer, not just for PI work.

If you’re a mid-market operator with a services line that AI is reshaping faster than your provider can keep up — CounselExpress is the shape of what CogleGroup builds. coglegroup.com is the parent firm. Calendar.

CounselExpress is live now at counselexpress.app. Tomorrow: ChamberAdvance — Cogle GTM 2/3.