The State of AI Maturity in Turkish Professional Services, 2026

Türkiye occupies a particular position in the European market for professional services. Through the 1995 Customs Union, the country sits inside the European Union's regulatory perimeter for goods but outside it for services and for most of the digital acquis, including the EU AI Act. Through commercial fact, a substantial share of the work performed in the country's larger law, accounting, consulting, and financial advisory firms is performed for clients headquartered in the Union, and to standards those clients export. This sector note, the first annual analysis the Institute publishes, examines how that position is shaping the adoption of artificial intelligence in the professional services market in Türkiye. The headline finding is that the sector clusters more tightly than comparable European markets at the lower-middle of the Standard, broadly AI-Aware shading into AI-Enabled, and that the constraints holding firms there are predominantly economic and organisational rather than technical or regulatory.

§ I Population and Method

This is the first iteration of an annual exercise the Institute intends to maintain. Between September 2025 and March 2026, the Institute reviewed adoption practices at 140 professional services firms in Türkiye, comprising 62 law firms, 41 accounting and audit practices, 24 management and strategy consultancies, and 13 financial advisory firms. The review combined structured interviews with managing partners, heads of risk, and chief operating officers; documentary inspection of internal policies, training records, and engagement letters where they bear on the use of deployed AI tools; and a structured survey of fee-earners regarding the foundation-model-based and other AI systems in use in matter execution. The methodology is published separately. No certification decisions are made on the basis of this exercise; the Institute's certifications proceed through the formal assessment process described in the methodology of the Standard. The population is not random. It is weighted to firms of more than 25 fee-earners, to firms with international client portfolios, and to firms headquartered in Istanbul, Ankara, and İzmir. The sampling biases are acknowledged and will be partially addressed in subsequent iterations.

§ II Six Findings

Six findings emerged from the review. Each is set out below as a single paragraph, in descending order of the Institute's confidence in the finding's generality.

First, the distribution against the tiers of the Standard clusters at the lower-middle and is shallowly tailed. Of the 140 firms reviewed, indicative evidence supported AI-Aware at approximately 38 percent, AI-Enabled at 44 percent, AI-Integrated at 12 percent, and a position approaching AI-First in one or two dimensions, though certifying lower on the weakest, at 6 percent. No firm reviewed presented evidence supporting AI-Native, and the Institute did not expect any to do so at this stage of the market's development. The clustering is consistent with patterns reported from comparable European jurisdictions one to two years into broad adoption of foundation-model-based tools, though it is more concentrated in Türkiye, with thinner tails. The distribution does not differ materially by practice area, with the partial exception of management consulting, which sits modestly higher on average and shows greater within-firm variation between practice groups.

Second, governance lags adoption almost universally. The most consistent observation across the 140 firms, true at every tier except the small group at the top of the distribution, is that the deployment of AI tools has run ahead of the firm's governance of them. Tools are in widespread use; written policies on their use are present in roughly 60 percent of firms reviewed but operative, in the sense of being known to fee-earners, applied in matter execution, and supported by training calibrated to roles, in fewer than 25 percent. Measurement is weaker still: fewer than 15 percent of firms reviewed could produce records of which deployed AI tools were used on which matters, by whom, with what supervision. This is the dimension on which most firms in the population will, when they come to formal assessment, certify at their weakest floor.

Tools are in widespread use; written policies on their use are operative in fewer than one firm in four.

Third, KVKK considerations and broader data-sovereignty concerns are reshaping vendor selection but not workflow design. The Turkish Personal Data Protection Law (Law No. 6698, as amended in 2024 by Law No. 7499) introduced an explicit framework for cross-border transfers that, while convergent with the General Data Protection Regulation in many respects, retains its own consent and adequacy provisions and is enforced by the Personal Data Protection Authority (Kişisel Verileri Koruma Kurumu)¹. Firms have responded by preferring domestic or EU-domiciled deployments of AI tools, and by negotiating data-residency provisions in vendor agreements. This is a substantive response at the procurement layer. It has not, in most firms, produced a corresponding response at the workflow layer: the question of how the deployed AI tool integrates with the firm's process, and whether the locus of professional judgement has shifted, is regularly conflated with the question of where the tool's data resides. The two are distinct, and the third dimension of the Standard is not satisfied by an answer that addresses only the second.

Fourth, partner-track economics actively disincentivise workflow redesign in a way that international commentary on professional services adoption has not adequately registered. In firms whose compensation models remain dominated by billable hours and a leverage pyramid in which junior fee-earners are the principal source of margin, the redesign of workflows around AI tools, which reduces the hours billed and, in some configurations, the number of junior fee-earners required, directly threatens the partnership's distributable profit. The structure is not unique to Türkiye, but it is more pronounced here, where partner cohorts in the larger firms are smaller and where the firm's economic dependence on a particular generation of senior partners is correspondingly higher. The result is that augmentation is widely embraced, since it improves margin without disturbing structure, and redesign is widely resisted, since it disturbs the structure on which the partnership rests. The point is made in the abstract in most strategy discussions in the firms reviewed; it is reflected, in the concrete, in how the firms have actually deployed the tools.

Fifth, pressure from European Union-based corporate clients is the single largest accelerant in the sector. Approximately a quarter of the firms reviewed, concentrated in international arbitration, cross-border mergers and acquisitions, banking and finance, and audit work for multinational groups, reported direct requests from EU-domiciled clients for evidence of AI literacy programmes under EU AI Act, Article 4, for representations regarding the firm's deployment of AI tools in the client's matters, or for participation in the client's third-party AI risk assessment. The pressure is not yet evenly distributed. It is concentrated in industries where the client is itself a deployer of high-risk AI systems within the meaning of EU AI Act, Annex III, and where the client's own compliance programme has reached the stage of supplier review. The pressure is, however, intensifying year on year, and the Institute expects it to extend to mid-market and domestic-facing practice within twelve to eighteen months.

Sixth, and least often noted by the firms themselves, the practical absence of legacy on-premise IT infrastructure is a structural advantage that the Turkish sector has not yet fully realised. The professional services firms reviewed are, on average, less encumbered by legacy document management systems, time-and-billing platforms, and bespoke practice software than their counterparts in jurisdictions with longer-standing investment in such infrastructure. Cloud-native architectures predominate, and the integration of deployed AI tools into matter execution is, in principle, technically more straightforward than in firms that must work around twenty-year-old document repositories. The advantage is real. It is not, in practice, being converted into a structural lead. The firms best placed to redesign their workflows are, in the main, the firms whose partnership economics most resist it. The technical conditions are favourable; the organisational conditions are not.

The technical conditions for workflow redesign are favourable in Türkiye. The organisational conditions are not.

§ III Cross-Cutting Observations

Two further observations did not rise to the level of findings but are reported here for their bearing on what the Institute expects to certify.

The first concerns the regulatory posture of the Banking Regulation and Supervision Agency (BDDK) and, in a more limited way, the Capital Markets Board (SPK). The BDDK's existing framework on outsourcing and on the use of cloud services in regulated financial institutions has acted, in practice, as a proxy for the governance of deployed AI systems among firms whose principal clients are regulated entities. Firms that have already documented their use of cloud services for banking clients have, almost as a side-effect, produced documentation that meets a substantial portion of what the Institute looks for at the second dimension of the Standard. This is an accelerant the sector should recognise more explicitly. It is also a reminder that the governance of artificial intelligence in Türkiye is being constructed less by a single horizontal regulation than by the cumulative effect of vertical supervisory expectations applied to particular sectors. Firms outside the regulated financial perimeter are correspondingly less advanced.

The second concerns talent. The Institute heard, in nearly every interview, some version of the observation that the country produces excellent technologists who do not, in general, choose to work in professional services firms. This is true in many jurisdictions, but the gap is wider in Türkiye, and the consequence is that the technology, knowledge management, and AI literacy functions within firms are thinner than the firms' size or ambitions would suggest. The Institute expects this to remain a constraint on the rate at which firms can move from AI-Enabled to AI-Integrated, irrespective of partner appetite. We do not propose remedies in this note; we note the constraint.

§ IV What the Institute Expects to Certify

The Institute opens formal certification under the AI Maturity Standard in Türkiye in the second quarter of 2026, with the first cohort of certified firms expected to be announced in the fourth quarter and a second cohort in the first half of 2027. Drawing on the population reviewed in this exercise, and recognising that the Institute does not invite or solicit certification (firms apply, and assessment proceeds on the firm's own application), the Institute's working expectation is as follows.

A first cohort of between twenty and thirty-five firms is plausible across 2026 and 2027, drawn predominantly from law and audit practice, with smaller representation from consulting and financial advisory. Within that cohort, the modal certification will be AI-Enabled. A meaningful minority, the Institute's working estimate is four to nine firms across the two-year period, will certify at AI-Integrated. At most one or two AI-First certifications are expected in this period, and only where a firm has done substantive work across all seven dimensions, not merely strong work in one. AI-Native certifications are not expected.

In parallel, the Institute expects to issue EU AI Act Readiness Attestations to a larger group of firms, perhaps fifty to eighty across 2026 and 2027, including firms whose AI Maturity Standard certification is at AI-Aware. The Attestation is calibrated to a regulatory threshold, not to a maturity assessment; a firm that has done thorough work on Article 4 literacy, on its inventory of deployed AI systems, and on the contractual and operational documentation that supports cross-border service to EU-based clients can satisfy the Attestation while sitting at any tier of the Standard. The Institute expects demand for the Attestation to outpace demand for full Standard certification through 2027, and regards that as the right ordering for the sector at this point in its development.

§ V The Institute's Position

The institutional position is restrained. The Turkish professional services sector is, at this stage, an early-middle adopter of artificial intelligence as a matter of operations: not the late laggard occasionally implied in international commentary, but not the leader some domestic accounts have suggested. The conditions for movement are present. The structural impediments are real and resistant to short-term remediation: governance lag, partner-track economics, talent thinness. The opportunities are real and available: client pressure from the Union, the absence of legacy infrastructure, sector-specific regulatory templates from the BDDK. Whether the sector closes the gap to its European counterparts over the next three years is, on the Institute's reading, less a question of capability than of partnership decision. The Institute will return to this question, with a second year of data, in 2027.