A proposal for Lynxeye · May 2026

"Accelerate Lynxeye with AI"

Accelerate Lynxeye with AI is about growing client impact by building on company AI infrastructure. HQ and Eraneos create the secure foundation. This work helps Lynxeye turn that foundation into smoother daily use, stronger client delivery, sharper pitches, and reusable practice.

Value matrix

Where the work creates value.

This is the agreement model: the visible table keeps the conversation simple, and the row details open the discussion points behind each level of work.

Level	Setup & daily use	Client projects	Client pitches
EmployeesIndividual fluency and confidence.	Setup & daily useSkills, tools, access, confidence	Client projectsBetter prepared, faster contribution	Client pitchesMore useful pitch input earlier
Discussion points	Setup & daily use Build confidence with approved tools, practical AI skills, and daily habits. Remove small frictions around files, context, voice input, browser work, and connectors. Help each person know when to use chat, when to use structured prompts, and when to ask for help.	Client projects Prepare better for meetings, research, synthesis, notes, and follow-up. Use AI for first drafts and alternatives while keeping judgment close to the work. Turn useful examples into reusable practice instead of one-off chat output.	Client pitches Contribute earlier with research, structure, options, and sharper synthesis. Reduce blank-page friction without reducing senior review. Capture what good pitch support looks like so others can repeat it.
TeamsShared working rhythm.	Setup & daily useShared setup and working rhythm	Client projectsWorkflows tested in real projects	Client pitchesBetter pitch flow and output quality
Discussion points	Setup & daily use Agree team patterns for briefing AI, storing context, sharing examples, and asking for review. Connect internal AI specialists, domain experts, and the pilot group around one practical rhythm. Make current wins, failures, and experiments visible so the team starts from reality.	Client projects Select 3 to 5 practical use cases with visible value. Test workflows in actual client work and prototypes, not generic training examples. Capture what worked, what failed, and what should become a Lynxeye method.	Client pitches Improve pitch preparation, proposal logic, research, and output formats. Use prototypes where they make the opportunity easier to see. Keep partner and senior review as part of the workflow from the start.
LynxeyeCompany-owned capability.	Setup & daily useCompany-owned adoption path	Client projectsPlaybooks and reusable methods	Client pitchesValue evidence and scale choices
Discussion points	Setup & daily use Create a practical setup guide for people beyond the pilot. Clarify approved tools, connectors, permissions, security, and daily operating habits. Make adoption a company capability, not a set of individual experiments.	Client projects Capture playbooks, guardrails, and reusable methods from real work. Map the next use cases and the bottlenecks AI can actually break. Decide what should scale, what needs more proof, and what should be left alone.	Client pitches Refine the value model with Lynxeye evidence. Turn the strongest pitch patterns into reusable methods and assets. Create a scale decision package that leadership can discuss with confidence.

The overview keeps the model simple: people build fluency, teams prove workflows, and Lynxeye turns the learning into reusable capability.

Level	Setup & daily use	Client projects	Client pitches
EmployeesWhat each person needs to start well.	Setup & daily useRemove personal friction	Client projectsFind useful work moments	Client pitchesMap pitch contribution
Discussion points	Setup & daily use Skills setup by role and current AI maturity. Approved AI tools configured correctly for each person. Projects, memory, custom instructions, files, and connectors mapped. Cloud storage, document access, and context hygiene clarified. OpenAI Atlas browser readiness where approved: bookmarks, page visibility, site permissions, and safe on-page help. Voice input, Mac, PC, browser, and phone friction removed where useful. Support needs and confidence barriers made visible.	Client projects Identify where each person can contribute faster in real client work. Map meeting preparation, transcript use, and follow-up friction. Find research, synthesis, documentation, translation, and versioning moments. Clarify what can be drafted, what can be reused, and what needs review. Confirm client data boundaries before any live workflow is tested.	Client pitches Map where individuals can prepare better for pitches. Identify moments for faster research, synthesis, structure, and options. Capture proposal, deck, and meeting-output friction. Clarify what support helps people contribute earlier. Decide what good pitch assistance should look like before senior review.
TeamsWhat the pilot should test together.	Setup & daily useUnderstand current practice	Client projectsSelect 3 to 5 use cases	Client pitchesSelect pitch contexts
Discussion points	Setup & daily use Current AI practice working session with internal AI specialists, domain experts, and pilot people. What works, what failed, what should be reused, and what should stop. Relevant HQ and group resources collected. Bridge team roles, actions, availability, and support rhythm clarified. Tool, connector, browser, storage, and approved workflow constraints mapped.	Client projects Rank 3 to 5 priority use cases that are visible, safe, and practical. Select project contexts with short-term value potential. Map team collaboration flows around AI. Identify prototype opportunities using Lynxeye materials. Decide what should stay chat, become structured chat, become automation, or become a small HTML tool. Define what evidence should be captured.	Client pitches Baseline the current pitch workflow. Select priority pitch contexts and growth moments. Identify reusable proposal logic and design elements. Map team review and decision points. Define pitch examples that show orchestration over ad hoc chat.
LynxeyeWhat leadership needs to agree.	Setup & daily useSet the operating boundaries	Client projectsDefine pilot scope	Client pitchesDefine value evidence
Discussion points	Setup & daily use HQ and June 3 input review. Tool, platform, security, data, IP, and compliance constraint map. Engagement charter, escalation paths, and decision rights. Clear rules for read, draft, create, edit, and send actions. Decision on whether Atlas or other browser-based AI workflows fit the approved environment.	Client projects Project Phase scope recommendation. Use cases, team, cadence, boundaries, and responsibilities. Baseline success measures and evidence plan. Compliance model for project material. Pricing basis for the next step, with pricing still [TBD].	Client pitches Baseline commercial simulator. Value capture assumptions. Leadership questions to resolve before the Project Phase. Evidence model for short-term value. What would make Lynxeye a forerunner inside the wider group.

Scope & Setup is where the work becomes concrete enough to agree: use cases, tools, people, evidence, boundaries, and what should be built first.

Level	Setup & daily use	Client projects	Client pitches
EmployeesPractice in the workday.	Setup & daily useHands-on setup and skills	Client projectsCoach real tasks	Client pitchesSupport better first drafts
Discussion points	Setup & daily use One-on-one practical setup support. Personalized tool environments, role context, and context packs. Connectors, storage, voice, and device habits working in daily use. Atlas/browser habits where approved: page context, permissions, bookmarks, and safe research. Chat routines turned into reusable prompts and working instructions. Meeting transcript to output routines.	Client projects Better meeting preparation. Faster documentation and follow-up. More useful first versions of client outputs. Reviewed outputs people can actually use. Clear notes on when AI should not be used. Learning captured from mistakes, drift, loops, and almost-right answers.	Client pitches Faster pitch preparation. More options and stronger first drafts. Pitch meeting notes turned into usable outputs. Research, synthesis, structure, and drafting support. Human judgment explicit before senior review.
TeamsLive workflows and prototypes.	Setup & daily useBuild shared routines	Client projectsRun workflows and prototypes	Client pitchesImprove pitch flow
Discussion points	Setup & daily use Shared setup standards. Storage hygiene and context patterns. Team support channel and action rules. Show-and-tell rhythm. Availability rhythm that keeps the work moving. Reusable examples of skills, connectors, and browser workflows that people can copy.	Client projects Designed workflows for selected use cases. Prototype builds using Lynxeye playbooks and design elements. Small HTML tools or micro-app prototypes where chat is too loose for repeated use. Team collaboration flows captured as deliverables. Before and after examples. Review and refine loop.	Client pitches Redesigned pitch workflow. Reusable proposal logic. Output formats and design patterns. Pitch prototypes and example outputs. Quality gates before external use. Clear path from prompt to draft to deck or HTML prototype when useful.
LynxeyeOutputs the company can reuse.	Setup & daily useDocument wider adoption	Client projectsCapture playbooks and use cases	Client pitchesPackage scale decision
Discussion points	Setup & daily use Documented setup guide for wider rollout. Interactive documentation hub. Tool setup, connector guidance, skills, browser habits, playbooks, and support material in one place. Ownership model for maintenance. Reskilling rhythm for the wider firm.	Client projects First Lynxeye playbooks. Use-case map and opportunity surface. Quality guardrails and review rhythm. Reusable methods owned by Lynxeye. What to scale and what to leave alone. Workflow patterns that can become automations or micro-apps over time.	Client pitches Refined simulator with Lynxeye evidence. Updated commercial upside quantification. Highest-value expansion recommendation. Staged scale plan. Lynxeye positioned as the go-to AI adoption and value creation reference inside the wider group.

The Project Phase is where the discussion points become working outputs: coached use, prototypes, playbooks, use-case evidence, and a clear next decision.

Level	Setup & daily use	Client projects	Client pitches
EmployeesConfidence, safety, and review.	Setup & daily useMake approved use practical	Client projectsKeep human review visible	Client pitchesProtect tone and quality
Discussion points	Setup & daily use Confidence and comfort checks. Support for skeptical or unsure employees. Approved tool use made practical. Consent rules for recording and transcripts. Atlas/browser page visibility, site permissions, and sensitive page handling where relevant. Clear escalation path for questions.	Client projects Human review for client-facing work. Safe handling of client data. Quality checks before output leaves the team. Notes on mistakes, weak outputs, and learning moments. QA sessions where people can ask practical questions.	Client pitches Review before pitch material is shared. Tone, quality, and originality checks. Safe use of client and prospect context. Clear rule for what can be drafted, reused, or shown. Better confidence before senior or partner review.
TeamsShared learning without blame.	Setup & daily useRun a team learning rhythm	Client projectsBuild QA into workflows	Client pitchesImprove patterns over time
Discussion points	Setup & daily use Every-other-week AI use-case session. Shared learning from wins, mistakes, and examples. Recognition for teams moving adoption forward. Practical AI team of the month or similar light ritual. Team support and Q&A cadence.	Client projects QA rhythm for live workflows. Escalation for privacy, IP, or quality concerns. Best practices captured from actual project work. Mistakes and lessons captured without blame. Review patterns added back into playbooks.	Client pitches Pitch quality review rhythm. Shared examples of strong AI-supported pitch work. Reusable pitch patterns improved over time. Team learning from what worked and what did not. Clear owner for maintaining pitch methods.
LynxeyeTrusted company capability.	Setup & daily useGovern the operating system	Client projectsMaintain reusable methods	Client pitchesFeed evidence back into growth
Discussion points	Setup & daily use Rhythm of business for AI. Governance for tools, connectors, browser use, permissions, and budgets. Regular capture of skills, use cases, best practices, mistakes, and learnings. Ownership for documentation and playbook updates. Wider adoption rhythm beyond the bridge team.	Client projects Company quality guardrails. Compliance, privacy, and IP habits. Reusable project methods maintained over time. Leadership visibility into adoption and value. A practical model for scaling safely.	Client pitches Commercial evidence reviewed regularly. Pitch and growth learnings fed back into the method. Strongest use cases prioritized for expansion. Recognition of teams creating measurable value. Lynxeye becomes a credible internal showcase for AI adoption.

Adoption, Security & Quality makes the work safe enough to repeat and strong enough for Lynxeye's premium standard.

Back of the envelope

Potential business value for Lynxeye.

This is not proof. It is a simple way to discuss the size of the opportunity: capacity captured, pitch speed, win probability, hiring pressure, and an editable investment frame. Scope & Setup calibrates the assumptions. Prove In Work replaces them with Lynxeye evidence.

Lynxeye baseline (SEK)

Employees 70

Assumption: Lynxeye headcount in scope for the model. Default: 70 from the brief. Range: 20 to 150, so the same logic can be tested on a smaller pilot or wider firm scale. Caveat: Phase 1 should confirm the real team count for Phase 2.

Billable hours / consultant / year 1,500

Assumption: annual billable hours per consultant. Default: 1,500, a directional premium consultancy starting point. Range: 1,000 to 2,000 for lower and higher utilization years. Caveat: Lynxeye-specific utilization should be confirmed before Phase 2.

Avg. billing rate (SEK/hr) 3,000

Assumption: blended billing rate across seniority levels. Default: 3,000 SEK/hr. Range: 1,500 to 5,000 SEK/hr because rates vary by role and engagement type. Caveat: this creates gross equivalent value, not net profit.

Improvement levers

Project delivery compression 25%

Assumption: reduction in project delivery time from better AI orchestration. Default: 25%, so a 12-week project becomes roughly 9 weeks. Range: 0 to 50%, from no lift to an ambitious upside case. Caveat: this must be proven by Lynxeye workflow evidence.

Pitch cycle compression 25%

Assumption: reduction in time from opportunity or RFP to proposal-ready. Default: 25%, so a 4-week pitch becomes roughly 3 weeks. Formula effect: baseline pitches divided by the compressed cycle. Caveat: baseline pitch volume is a placeholder.

Win rate improvement 15%

Assumption: relative improvement on qualified pitch wins from better preparation and faster turnaround. Default: 15%. Range: 0 to 40%. Caveat: this affects expected additional wins only, no revenue value is assigned because deal size is not in the model.

% of freed time captured as real value 60%

Assumption: share of freed time that becomes real value instead of work creep. Default: 60%. Formula effect: raw freed capacity is multiplied by this rate. Caveat: Phase 1 defines how this will be measured, and Phase 2 tests it in real work.

Tooling cost note. There will of course be a cost per employee for AI solutions and tools. This simulator keeps that outside the model until the exact stack, licenses, and scope are clear.

Advanced assumptions

Qualified pitches / year Assumption: qualified pitch volume before compression. Default: 12, roughly one qualified pitch per month. Formula effect: used in pitch-equivalent capacity and expected wins. Caveat: Henrik should replace this with Lynxeye's actual volume.

Baseline win rate Assumption: current win rate on qualified pitches. Default: 35%. Formula effect: baseline pitches x baseline win rate x win-rate improvement. Caveat: this is not Lynxeye-confirmed, and revenue impact is intentionally excluded.

Working weeks / year Assumption: working weeks used for human interpretation. Default: 46. Formula effect: captured capacity divided by employees and working weeks. Caveat: this is only for readability, not the core commercial value calculation.

Uncertainty band Assumption: uncertainty band around captured capacity and gross value. Default: 15%. Formula effect: captured capacity is shown as plus or minus this band. Caveat: exact internal math is rounded only for display.

Output · directional Phase 2 potential at current settings

Main signal

Captured capacity

Formula: raw freed capacity x freed-time capture rate, shown with the uncertainty band. Interpretation: the part of the upside that survives work creep. Caveat: this is the directional number Phase 2 is meant to test with evidence.

13,400 to 18,100 hours

For Lynxeye at 70 people, roughly 15,750 hours survives work creep before the uncertainty band.

Gross capacity value

Formula: captured capacity range x blended billing rate. Interpretation: gross equivalent capacity value if the time becomes useful client work. Caveat: this is not net profit, and it excludes margin, pricing, demand, and change cost.

40.2 to 54.3M SEK

If captured as billable work at 3,000 SEK/hr, this is the gross directional Phase 2 upside.

Pitch-equivalent capacity

Formula: baseline pitches divided by (1 minus pitch compression), then minus baseline pitches. Interpretation: extra pitch-equivalent capacity from shorter cycles. Caveat: this is capacity, not a sales forecast.

At 12 baseline pitches and 25% compression, the same pitch capacity can handle roughly +4 more pitch-equivalents per year.

Hours returned / consultant / week

Formula: captured capacity divided by employees and working weeks. Interpretation: what the model feels like in the week. Caveat: this depends on leaders protecting the recovered time from low-value work.

4.9 hours

Roughly 0.7 workdays per consultant per week.

Model context

Raw freed capacity

Formula: employees x billable hours x project delivery compression. Interpretation: gross time opportunity before capture. Caveat: raw freed time is not value until the organization captures it.

26,250 hours

Gross time opportunity before capture rate is applied.

Expected additional qualified wins

Formula: baseline pitches x baseline win rate x win-rate improvement. Interpretation: incremental qualified wins before any deal-size value. Caveat: this is directional only, no revenue value is assigned.

+0.6

At 12 qualified pitches per year and a 35% baseline win rate.

AI tool costs

Interpretation: there will be a cost per employee for AI solutions and tools. Caveat: the exact tool stack, licensing model, and enterprise terms should be handled after scope alignment, not inside this directional upside model.

Separate implementation cost.

Tooling is real, but it is not a slider in this model.

Captured vs. absorbed time

Formula: captured time equals raw freed capacity x capture rate, absorbed time is the remainder. Interpretation: the work-creep risk made visible. Caveat: the split is a hypothesis until measured.

60% captured / 40% absorbed

Phase 1 defines the split. Phase 2 is where it can move in the right direction.

Where the time can go

Interpretation: recovered effort can move into higher-value work and less blank-page production burden. Caveat: this is not measured yet, and the pattern depends on leadership rhythm.

Client prep, synthesis, senior review, learning, proposal depth.

A better use of the week, not just a shorter week.

Every 1 percentage point increase in captured time is worth roughly 263 hours / 788k SEK gross capacity at current settings.

Investment frame

Compare the directional value with an editable investment frame.

This is a value conversation tool, not a guarantee. Add a working investment frame to see what share of the projected value would need to be captured.

Investment to compare (SEK) Use this as an editable comparison frame, not a quoted fixed price. The simulator keeps gross directional value separate from any later offer decision.

Net value after investment Waiting for investment

Value multiple Waiting for investment

Payback signal Waiting for investment

Add an investment frame to compare it with the directional value range.

What this means

For clients

Directional interpretation. The simulator does not prove client outcomes. It shows where freed capacity could improve the client experience if Lynxeye captures it deliberately.

Faster iteration, better-prepared teams, and more senior attention on judgment, implications, and quality.

For employees

Directional interpretation. Employee value depends on whether leadership protects the recovered time from being absorbed by more low-value work.

Fewer blank-page starts, less production drag, clearer playbooks, and more time for thinking and craft.

For partners

Directional interpretation. The partner value is not that senior people disappear from the work. It is that their judgment is used at better moments.

Less senior bottleneck, better delegation, reusable methods, and clearer quality control across teams.

For Lynxeye

Directional interpretation. The firm-level value only appears when isolated AI use becomes shared operating capability.

Compounding capability instead of isolated tool use: workflows, playbooks, and judgment that improve with reuse.

Directional Phase 2 potential. These numbers are directional, not promises. The purpose is to frame what becomes possible once the pilot creates evidence.

Scope before proof

Before the upside can be claimed, Lynxeye needs a clean baseline and a focused pilot design.

Phase 1 defines what to measure, which teams to involve, which work to use, and what the central AI setup makes possible.

Phase 2 then tests the value in real client work, not in a theoretical business case.

Hiring pressure

Could capability lower the next hiring need?

This is not a headcount reduction argument. It is a hiring pressure argument: hire where human judgment is needed, and let playbooks absorb repeatable production load.

Pricing stays [TBD]

Hiring need under the old answer 3 hires

Assumption: the number of hires Lynxeye might consider if the only answer is more headcount. Default: 3, matching the finance-team pressure mentioned in the brief. Range: 1 to 6, so the pressure can be tested without changing the rest of the model.

Capability support level

50%

Scenario

Old answer: hire 3. Capability answer: hire about 2, while playbooks absorb the pressure of roughly 1 role.

Hiring run-rate not immediately added 1.2 to 1.5M SEK / year

AI tool costs Handled separately

Project Accelerate pricing stays [TBD], and this model always leaves at least one real hire in place.

Old hiring answer 3.6 to 4.5M SEK / year

Fully loaded annual run-rate before recruiting delay.

Capability answer About 2 hires still needed

Plus [TBD] Phase 1 setup, Phase 2 engagement, and per-employee AI tool costs outside the hiring-cost model.

Time to first lift 8-12 weeks

Hiring can still be right. Capability work can start easing pressure before a 6-12 month hiring cycle has fully landed.

What we do not know yet: actual adoption rate, true time savings per workflow, and how much repeatable load playbooks can absorb. Phase 1 defines how to test those numbers. Phase 2 turns them into evidence.

Where these numbers come from

Source status: directional references for the draft model. Henrik should confirm which sources to cite before external sharing.

Productivity potential: McKinsey, "The Economic Potential of Generative AI".

Task-level consulting evidence: HBS AI Institute, "Navigating the Jagged Technological Frontier".

Adoption and value capture: BCG on value creation and destruction with GenAI, BCG on capability expansion, and McKinsey on rewiring to capture value.

Hiring cost logic: Skatteverket employer contributions, used only to support fully loaded Swedish employer-cost logic.

Lynxeye-specific: conversations with Lynxeye, May 2026. Defaults are deliberately directional and should be calibrated in Phase 1.

Direction, not promise. The model frames the upside Phase 2 could help Lynxeye investigate and qualify. It does not claim that Phase 1 delivers these numbers by itself.